Audio Decision Model demo

.gitattributes

@@ -0,0 +1,62 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+examples/music-639.wav filter=lfs diff=lfs merge=lfs -text
+examples/nature-9508.wav filter=lfs diff=lfs merge=lfs -text
+examples/reservation.wav filter=lfs diff=lfs merge=lfs -text
+examples/animal-863.wav filter=lfs diff=lfs merge=lfs -text
+examples/irodori-reservation.wav filter=lfs diff=lfs merge=lfs -text
+examples/irodori-change.wav filter=lfs diff=lfs merge=lfs -text
+examples/irodori-cancel.wav filter=lfs diff=lfs merge=lfs -text
+fonts/LINESeedJP_OTF_Bd.woff2 filter=lfs diff=lfs merge=lfs -text
+fonts/LINESeedJP_OTF_Rg.woff2 filter=lfs diff=lfs merge=lfs -text
+examples/natural-reservation-a.wav filter=lfs diff=lfs merge=lfs -text
+examples/natural-change-a.wav filter=lfs diff=lfs merge=lfs -text
+examples/natural-cancel-b.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-order-full.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-backchannel-full.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-confirm-full.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-correction-full.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-cancel-full.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-order-hesitation-unfinished.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-reservation-clause-unfinished.wav filter=lfs diff=lfs merge=lfs -text
+examples/turn-reservation-full-contrast.wav filter=lfs diff=lfs merge=lfs -text
+examples/natural-reservation-a-telephone.wav filter=lfs diff=lfs merge=lfs -text
+examples/adult-reservation-v3.wav filter=lfs diff=lfs merge=lfs -text
+examples/adult-backchannel-v3.wav filter=lfs diff=lfs merge=lfs -text
+examples/adult-cancel-v3.wav filter=lfs diff=lfs merge=lfs -text
+examples/music-guitar-369676.wav filter=lfs diff=lfs merge=lfs -text
+examples/animal-dog-birds-238688.wav filter=lfs diff=lfs merge=lfs -text
+examples/music-acoustic-loop-848704.wav filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -0,0 +1,26 @@
+FROM python:3.12-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PORT=7860 \
+    HOME=/home/user \
+    HF_HOME=/home/user/.cache/huggingface
+
+WORKDIR /app
+COPY requirements.txt ./
+# The CPU wheel index prevents a CUDA runtime from being included in this
+# CPU-only Space.  The qwen-asr 0.0.6 transformer backend is vendored under
+# server_runtime, avoiding its unrelated ASR UI and alignment dependencies.
+RUN pip install --no-cache-dir --index-url https://download.pytorch.org/whl/cpu torch==2.9.1 && \
+    pip install --no-cache-dir -r requirements.txt
+# Pin both remote bases at build time. Runtime loading names exact revisions,
+# so a request cannot silently move to a different model version.
+RUN python -c "from huggingface_hub import snapshot_download; snapshot_download('Qwen/Qwen3-ASR-0.6B', revision='5eb144179a02acc5e5ba31e748d22b0cf3e303b0'); snapshot_download('laion/clap-htsat-unfused', revision='8fa0f1c6d0433df6e97c127f64b2a1d6c0dcda8a')"
+RUN useradd --create-home --uid 1000 user && \
+    mkdir -p /home/user/.cache/huggingface && \
+    chown -R user:user /home/user
+COPY --chown=1000:1000 . .
+
+USER user
+EXPOSE 7860
+CMD ["sh", "-c", "uvicorn server:app --host 0.0.0.0 --port ${PORT} --no-access-log"]

README.md

@@ -0,0 +1,33 @@
+---
+title: 音声版Jev「Audio Decision Model」
+emoji: 🎧
+colorFrom: indigo
+colorTo: blue
+sdk: docker
+app_port: 7860
+pinned: false
+---
+
+# 音声版Jev「Audio Decision Model」
+
+音声と質問を入力すると、`Noul`（Yes / No）・`Choice`（選択）・`Score`（数値）の構造化JSONで答えます。予約の取り消しやキーワード、音楽や動物の鳴き声など、発話内容と周囲の音について質問できます。
+
+質問と選択肢は画面上で編集できます。CloudはQwen3-ASR-0.6B、ModernBERT-Ja、CLAPを組み合わせて処理します。Edge AIは量子化したONNXモデルをブラウザ上で実行し、入力データをクラウドへ送りません。
+
+[モデル](https://huggingface.co/mocomoco-inc/mocomoco-AudioJev-AudioDecisionModel) · [mocomoco inc.](https://mocomoco.ai)
+
+## 使い方
+
+1. CloudまたはEdge AIを選びます。
+2. 30秒以内の音声を録音・アップロードするか、サンプルを選びます。
+3. 質問と選択肢を編集して「実行する」を押します。
+
+## 評価
+
+予約・注文などの用件と、依頼・確認など発話の役割を音声から同時に答える評価では、事前に固定した240件のうち**88.75%**で両方正解しました。発話と周囲の音を組み合わせる質問や、自由に作成した質問全体の正答率は引き続き評価中です。
+
+## ライセンス
+
+Qwen3-ASRとCLAPはApache-2.0、ModernBERT-JaはMITです。モデルに含まれるファイルごとの条件は[モデルのLICENSE.md](https://huggingface.co/mocomoco-inc/mocomoco-AudioJev-AudioDecisionModel/blob/main/LICENSE.md)に記載しています。デモ内のサンプル音声の権利表示は`examples/ATTRIBUTION.json`にあります。
+
+[Jev / TypeSafe AI](https://typesafe.ai/blog/introducing-system-one-models-and-jev)と[OpenJev / SemIf](https://github.com/TheoLeeCJ/SemIf)を参考にしています。

app.py

@@ -0,0 +1,321 @@
+"""Private Hugging Face Space UI for the Mocovoice Audio Jev research demo."""
+
+from __future__ import annotations
+
+import importlib
+import json
+import os
+from pathlib import Path
+from typing import Any
+
+import gradio as gr
+
+
+SPACE_ROOT = Path(__file__).resolve().parent
+CATALOG_PATH = SPACE_ROOT / "examples" / "catalog.json"
+
+DEFAULT_QUESTIONS = {
+    "本人確認": {
+        "type": "noul",
+        "instructions": "話者が本人確認に必要な情報を明確に述べているか確認してください。",
+        "criteria": {
+            "false": "必要な情報が不足している、または聞き取れません。",
+            "true": "必要な情報を明確に述べています。",
+        },
+    },
+    "聞き取りやすさ": {
+        "type": "score",
+        "instructions": "発話の明瞭さを評価してください。",
+        "criteria": ["聞き取りにくい", "一部聞き取りにくい", "概ね明瞭", "明瞭"],
+    },
+}
+
+PRESETS: dict[str, tuple[str, dict[str, Any]]] = {
+    "本人確認": (
+        "本人確認の会話です。氏名、生年月日など必要な情報が音声に含まれているか確認します。",
+        DEFAULT_QUESTIONS,
+    ),
+    "応対品質": (
+        "顧客応対の通話です。応対内容と話し方を評価します。",
+        {
+            "要件の確認": {
+                "type": "noul",
+                "instructions": "依頼・質問の要点を確認できているか評価してください。",
+                "criteria": {"false": "要点を確認できていません。", "true": "要点を確認できています。"},
+            },
+            "応対評価": {
+                "type": "score",
+                "instructions": "丁寧で分かりやすい応対か評価してください。",
+                "criteria": ["改善が必要", "やや不十分", "適切", "非常に良い"],
+            },
+        },
+    ),
+    "選択肢の分類": (
+        "この音声を、定義した選択肢のいずれかに分類します。",
+        {
+            "お問い合わせ種別": {
+                "type": "choice",
+                "instructions": "主なお問い合わせ内容を一つ選んでください。",
+                "criteria": {
+                    "契約": "契約内容、変更、解約に関する問い合わせ",
+                    "請求": "料金、請求、支払いに関する問い合わせ",
+                    "技術": "利用方法や不具合に関する問い合わせ",
+                },
+            }
+        },
+    ),
+}
+
+
+def _pretty(value: Any) -> str:
+    return json.dumps(value, ensure_ascii=False, indent=2)
+
+
+def _user_error(message: str) -> None:
+    raise gr.Error(message)
+
+
+def _parse_json(value: str, label: str, *, allow_empty: bool = False) -> Any:
+    if not value or not value.strip():
+        if allow_empty:
+            return None
+        _user_error(f"{label}を入力してください。")
+    try:
+        return json.loads(value)
+    except json.JSONDecodeError as exc:
+        _user_error(f"{label}は有効なJSONにしてください（{exc.msg}）。")
+
+
+def _validate_questions(raw: Any) -> dict[str, dict[str, Any]]:
+    if not isinstance(raw, dict) or not raw:
+        _user_error("質問JSONは、質問名をキーにした1件以上のオブジェクトにしてください。")
+    validated: dict[str, dict[str, Any]] = {}
+    for name, question in raw.items():
+        if not isinstance(name, str) or not name.strip():
+            _user_error("質問名は空でない文字列にしてください。")
+        if not isinstance(question, dict):
+            _user_error(f"「{name}」はオブジェクトにしてください。")
+        kind = question.get("type")
+        instructions = question.get("instructions")
+        criteria = question.get("criteria")
+        if kind not in {"noul", "choice", "score"}:
+            _user_error(f"「{name}」のtypeは noul / choice / score のいずれかにしてください。")
+        if not isinstance(instructions, str) or not instructions.strip():
+            _user_error(f"「{name}」のinstructionsを入力してください。")
+        if kind == "noul":
+            if criteria is None:
+                criteria = {"false": "該当しません。", "true": "該当します。"}
+            if not isinstance(criteria, dict) or set(criteria) != {"false", "true"}:
+                _user_error(f"「{name}」のnoul criteriaは false と true を持つオブジェクトにしてください。")
+        elif kind == "choice":
+            if not isinstance(criteria, dict) or not 2 <= len(criteria) <= 32:
+                _user_error(f"「{name}」のchoice criteriaは2〜32件の選択肢オブジェクトにしてください。")
+        else:
+            if not isinstance(criteria, list) or not 2 <= len(criteria) <= 10:
+                _user_error(f"「{name}」のscore criteriaは2〜10件の順序���きリストにしてください。")
+        validated[name.strip()] = {"type": kind, "instructions": instructions.strip(), "criteria": criteria}
+    return validated
+
+
+def _questions_from_text(questions_text: str) -> dict[str, dict[str, Any]]:
+    return _validate_questions(_parse_json(questions_text, "質問JSON"))
+
+
+def _criteria_hint(kind: str) -> str:
+    if kind == "noul":
+        return _pretty({"false": "該当しません", "true": "該当します"})
+    if kind == "choice":
+        return _pretty({"選択肢A": "説明", "選択肢B": "説明"})
+    return _pretty(["低い", "高い"])
+
+
+def _set_criteria_hint(kind: str) -> str:
+    return _criteria_hint(kind)
+
+
+def _edit_question(
+    questions_text: str, name: str, kind: str, instructions: str, criteria_text: str, action: str
+) -> tuple[str, str, str, str]:
+    questions = _questions_from_text(questions_text)
+    clean_name = name.strip()
+    if not clean_name:
+        _user_error("質問名を入力してください。")
+    if action == "remove":
+        if clean_name not in questions:
+            _user_error(f"「{clean_name}」は質問JSONにありません。")
+        del questions[clean_name]
+        if not questions:
+            _user_error("最後の質問は削除できません。代わりに内容を更新してください。")
+        return _pretty(questions), "", "", _criteria_hint("noul")
+
+    criteria = _parse_json(criteria_text, "基準") if criteria_text.strip() else None
+    candidate = {clean_name: {"type": kind, "instructions": instructions, "criteria": criteria}}
+    normalized = _validate_questions(candidate)[clean_name]
+    questions[clean_name] = normalized
+    return _pretty(questions), "", "", _criteria_hint(kind)
+
+
+def _load_preset(name: str) -> tuple[str, str]:
+    context, questions = PRESETS[name]
+    return context, _pretty(questions)
+
+
+def _catalog_examples() -> list[list[Any]]:
+    if not CATALOG_PATH.is_file():
+        return []
+    try:
+        records = json.loads(CATALOG_PATH.read_text(encoding="utf-8"))
+        examples: list[list[Any]] = []
+        for record in records:
+            if not all(key in record for key in ("name", "audio", "context", "questions")):
+                continue
+            audio = SPACE_ROOT / str(record["audio"])
+            if audio.is_file():
+                examples.append([str(audio), str(record["context"]), _pretty(record["questions"])])
+        return examples
+    except (OSError, json.JSONDecodeError, TypeError):
+        return []
+
+
+def _backend() -> Any:
+    try:
+        return importlib.import_module("demo_backend")
+    except Exception:
+        _user_error("デモの実行準備がまだ完了していません。少し待ってから再度お試しください。")
+
+
+def _summary(result: Any) -> str:
+    if not isinstance(result, dict):
+        return "結果を受け取りました。下のJSONで詳細を確認してください。"
+    lines = ["### 読みやすい結果"]
+    transcript = result.get("transcript") or result.get("text")
+    if isinstance(transcript, str) and transcript.strip():
+        lines.append(f"**文字起こし**  \\n{transcript.strip()}")
+    answers = result.get("answers") or result.get("results") or result.get("decisions")
+    if isinstance(answers, dict):
+        for name, answer in answers.items():
+            if isinstance(answer, dict):
+                value = answer.get("value", answer.get("answer", answer.get("label", answer)))
+                rationale = answer.get("reason") or answer.get("rationale")
+                lines.append(f"- **{name}**: {value}" + (f" — {rationale}" if rationale else ""))
+            else:
+                lines.append(f"- **{name}**: {answer}")
+    elif not transcript:
+        lines.append("下のJSONで結果を確認してください。")
+    return "\n".join(lines)
+
+
+def _analyze(audio_path: str | None, context: str, questions_text: str) -> tuple[Any, str]:
+    if not audio_path:
+        _user_error("音声ファイルをアップロードするか、マイクで録音してください。")
+    questions = _questions_from_text(questions_text)
+    try:
+        result = _backend().analyze(str(audio_path), context or "", _pretty(questions))
+    except gr.Error:
+        raise
+    except Exception:
+        _user_error("解析に失敗しました。音声と質問の内容を確認して、もう一度お試しください。")
+    return result, _summary(result)
+
+
+def _export_sample(
+    audio_path: str | None, context: str, questions_text: str, expected_text: str, result: Any
+) -> str:
+    if not audio_path:
+        _user_error("サンプルを保存するには音声を指定してください。")
+    questions = _questions_from_text(questions_text)
+    expected = _parse_json(expected_text, "期待する回答", allow_empty=True)
+    result_json = result if isinstance(result, dict) else None
+    try:
+        return _backend().export_sample(str(audio_path), context or "", _pretty(questions), _pretty(expected) if expected is not None else "", result_json)
+    except gr.Error:
+        raise
+    except Exception:
+        _user_error("サンプルZIPを作成できませんでした。入力内容を確認して再度お試しください。")
+
+
+def _model_status() -> str:
+    try:
+        return str(_backend().model_status())
+    except gr.Error:
+        return "準備状況を確認できません。"
+    except Exception:
+        return "準備状況を確認できません。"
+
+
+def build_app() -> gr.Blocks:
+    css = """
+    .gradio-container { max-width: 1120px !important; }
+    #hero { padding: 1.2rem 0 .4rem; }
+    #notice { border-left: 4px solid #e0a42b; padding: .7rem 1rem; background: #fff8e8; border-radius: 6px; }
+    """
+    with gr.Blocks(title="Mocovoice Audio Jev", css=css) as demo:
+        gr.Markdown("# Mocovoice Audio Jev", elem_id="hero")
+        gr.Markdown(
+            "音声と文脈から、定義した質問への回答を試す研究プロトタイプです。"
+            "新しい業務・質問への精度は検証されていません。結果は確認用であり、自動処理は行いません。",
+            elem_id="notice",
+        )
+        with gr.Row():
+            with gr.Column(scale=1):
+                audio = gr.Audio(label="音声", sources=["upload", "microphone"], type="filepath")
+                context = gr.Textbox(label="文脈・補足", lines=5, placeholder="例：この通話は契約変更についてのお問い合わせです。")
+                preset = gr.Dropdown(label="プリセット", choices=list(PRESETS), value="本人確認")
+                load_preset = gr.Button("プリセットを読み込む")
+            with gr.Column(scale=1):
+                questions = gr.Textbox(label="質問JSON", value=_pretty(DEFAULT_QUESTIONS), lines=18)
+                with gr.Accordion("質問をフォームで編集", open=False):
+                    gr.Markdown("質問名が同じ場合は更新します。削除は質問名を指定して行います。")
+                    question_name = gr.Textbox(label="質問名")
+                    question_type = gr.Radio(["noul", "choice", "score"], value="noul", label="回答タイプ")
+                    question_instructions = gr.Textbox(label="指示", lines=3)
+                    question_criteria = gr.Textbox(label="基準JSON", value=_criteria_hint("noul"), lines=5)
+                    with gr.Row():
+                        add_update = gr.Button("追加・更新", variant="secondary")
+                        remove = gr.Button("削除", variant="stop")
+
+        with gr.Row():
+            analyze = gr.Button("解析する", variant="primary", scale=1)
+            status_button = gr.Button("準備状況を確認", scale=0)
+        status = gr.Textbox(label="モデルの準備状況", interactive=False, value="解析時にモデルを読み込みます。")
+
+        with gr.Row():
+            with gr.Column():
+                summary = gr.Markdown("解析すると、ここに読みやすい結果を表示します。")
+            with gr.Column():
+                result = gr.JSON(label="結果JSON")
+
+        with gr.Accordion("評価サンプルを保存", open=False):
+            gr.Markdown("解析を待たずに、音声・文脈・質問をデータセット用ZIPとして保存できます。期待する回答は任意です。")
+            expected = gr.Textbox(label="期待する回答JSON（任意）", lines=7, placeholder='例：{"本人確認": true}')
+            export = gr.Button("サンプルZIPを作成")
+            download = gr.File(label="サンプルZIP")
+
+        examples = _catalog_examples()
+        if examples:
+            gr.Examples(examples=examples, inputs=[audio, context, questions], label="収録済みの例")
+
+        load_preset.click(_load_preset, inputs=preset, outputs=[context, questions])
+        question_type.change(_set_criteria_hint, inputs=question_type, outputs=question_criteria)
+        add_update.click(
+            lambda q, n, t, i, c: _edit_question(q, n, t, i, c, "upsert"),
+            inputs=[questions, question_name, question_type, question_instructions, question_criteria],
+            outputs=[questions, question_name, question_instructions, question_criteria],
+        )
+        remove.click(
+            lambda q, n, t, i, c: _edit_question(q, n, t, i, c, "remove"),
+            inputs=[questions, question_name, question_type, question_instructions, question_criteria],
+            outputs=[questions, question_name, question_instructions, question_criteria],
+        )
+        analyze.click(_analyze, inputs=[audio, context, questions], outputs=[result, summary])
+        status_button.click(_model_status, outputs=status)
+        export.click(_export_sample, inputs=[audio, context, questions, expected, result], outputs=download)
+    return demo.queue(default_concurrency_limit=1)
+
+
+if __name__ == "__main__":
+    build_app().launch(
+        server_name="0.0.0.0",
+        server_port=int(os.environ.get("PORT", "7860")),
+        share=False,
+    )

audio-player.js

@@ -0,0 +1,133 @@
+const playIcon = `<svg aria-hidden="true" viewBox="0 0 24 24" fill="currentColor"><path d="M8 5.3v13.4c0 .8.9 1.3 1.6.8l10.1-6.7a1 1 0 0 0 0-1.6L9.6 4.5A1 1 0 0 0 8 5.3Z"/></svg>`;
+const pauseIcon = `<svg aria-hidden="true" viewBox="0 0 24 24" fill="currentColor"><rect x="6" y="5" width="4.5" height="14" rx="1"/><rect x="13.5" y="5" width="4.5" height="14" rx="1"/></svg>`;
+
+function formatTime(value) {
+  const seconds = Number.isFinite(value) ? Math.max(0, Math.floor(value)) : 0;
+  return `${Math.floor(seconds / 60)}:${String(seconds % 60).padStart(2, "0")}`;
+}
+
+function userError(error) {
+  if (error?.name === "NotAllowedError") return "再生がブラウザにより停止されました。再生ボタンをもう一度押してください。";
+  if (error?.name === "NotSupportedError") return "この音声形式はブラウザで再生できません。別の音声ファイルをお試しください。";
+  return "音声を再生できませんでした。音声ファイルを確認して、もう一度お試しください。";
+}
+
+/** Controls a native audio element and draws decoded mono 16 kHz samples locally. */
+export function createAudioPlayer({ audio, container, canvas, playButton, seek, time, speed, onError } = {}) {
+  if (!audio || !container || !canvas || !playButton || !seek || !time || !speed) {
+    throw new Error("音声プレーヤーのUI要素が不足しています。");
+  }
+
+  const context = canvas.getContext("2d");
+  let samples = new Float32Array();
+  let animationFrame = 0;
+  let lastDrawAt = 0;
+  let disposed = false;
+  let resizeObserver = null;
+
+  canvas.setAttribute("role", "img");
+  canvas.setAttribute("aria-label", "音声の波形と再生位置");
+  seek.min = "0"; seek.step = "0.01"; seek.value = "0";
+  seek.setAttribute("aria-label", "再生位置");
+
+  function duration() { return Number.isFinite(audio.duration) ? audio.duration : 0; }
+  function setState(value) { container.dataset.playerState = value; }
+  function setButton(playing) {
+    playButton.setAttribute("aria-label", playing ? "一時停止" : "再生");
+    playButton.setAttribute("aria-pressed", String(playing));
+    playButton.innerHTML = playing ? pauseIcon : playIcon;
+  }
+  function updateText() { time.textContent = `${formatTime(audio.currentTime)} / ${formatTime(duration())}`; }
+  function updateSeek() { seek.max = String(duration()); seek.value = String(Math.min(duration(), audio.currentTime || 0)); updateText(); }
+
+  function canvasSize() {
+    const rect = canvas.getBoundingClientRect();
+    const dpr = window.devicePixelRatio || 1;
+    const width = Math.max(1, Math.round(rect.width * dpr));
+    const height = Math.max(1, Math.round(rect.height * dpr));
+    if (canvas.width !== width || canvas.height !== height) { canvas.width = width; canvas.height = height; }
+    return { width, height, dpr };
+  }
+
+  function draw() {
+    const { width, height, dpr } = canvasSize();
+    context.clearRect(0, 0, width, height);
+    const style = getComputedStyle(container);
+    const brand = style.getPropertyValue("--brand").trim() || "#7203ff";
+    const muted = style.getPropertyValue("--wave-muted").trim() || "#c6c8d0";
+    const binCount = Math.max(24, Math.min(140, Math.round((width / dpr) / 4)));
+    const progress = duration() ? Math.max(0, Math.min(1, audio.currentTime / duration())) : 0;
+    const gap = Math.max(1, Math.round(dpr));
+    const barWidth = Math.max(1, (width - gap * (binCount - 1)) / binCount);
+    const minHeight = Math.max(2 * dpr, height * 0.07);
+    for (let bin = 0; bin < binCount; bin += 1) {
+      const start = Math.floor(bin * samples.length / binCount);
+      const end = Math.max(start + 1, Math.floor((bin + 1) * samples.length / binCount));
+      let peak = 0;
+      for (let index = start; index < end; index += 1) peak = Math.max(peak, Math.abs(samples[index] || 0));
+      const barHeight = Math.max(minHeight, peak * (height * 0.9));
+      const x = bin * (barWidth + gap);
+      context.fillStyle = (bin + 0.5) / binCount <= progress ? brand : muted;
+      context.beginPath();
+      if (typeof context.roundRect === "function") context.roundRect(x, (height - barHeight) / 2, barWidth, barHeight, Math.min(barWidth / 2, 3 * dpr));
+      else context.rect(x, (height - barHeight) / 2, barWidth, barHeight);
+      context.fill();
+    }
+  }
+
+  function animate(now) {
+    if (audio.paused || disposed) { animationFrame = 0; return; }
+    animationFrame = requestAnimationFrame(animate);
+    if (now - lastDrawAt >= 50) { lastDrawAt = now; updateSeek(); draw(); }
+  }
+  function startAnimation() {
+    if (!animationFrame) { lastDrawAt = 0; animationFrame = requestAnimationFrame(animate); }
+  }
+  function stopAnimation() { cancelAnimationFrame(animationFrame); animationFrame = 0; }
+
+  async function togglePlay() {
+    if (!audio.paused) { audio.pause(); return; }
+    setState("waiting"); playButton.disabled = true;
+    try { await audio.play(); }
+    catch (error) { setState("error"); setButton(false); onError?.(new Error(userError(error))); }
+    finally { playButton.disabled = false; }
+  }
+
+  function onMetadata() {
+    container.hidden = false; updateSeek(); setButton(false); setState("ready"); draw();
+  }
+  function onSourceReset() {
+    stopAnimation(); seek.value = "0"; time.textContent = "0:00 / 0:00"; setButton(false); setState("loading"); draw();
+  }
+  function onPlay() { setState("playing"); setButton(true); startAnimation(); }
+  function onPause() { setState("paused"); setButton(false); stopAnimation(); updateSeek(); draw(); }
+  function onEnded() { setState("ended"); setButton(false); stopAnimation(); updateSeek(); draw(); }
+  function onWaiting() { setState("waiting"); }
+  function onErrorEvent() { setState("error"); setButton(false); stopAnimation(); onError?.(new Error(userError(audio.error))); }
+
+  const listeners = [
+    ["loadedmetadata", onMetadata], ["durationchange", updateSeek], ["loadstart", onSourceReset], ["emptied", onSourceReset],
+    ["timeupdate", () => { updateSeek(); if (audio.paused) draw(); }], ["play", onPlay], ["pause", onPause], ["ended", onEnded], ["waiting", onWaiting], ["error", onErrorEvent]
+  ];
+  listeners.forEach(([event, handler]) => audio.addEventListener(event, handler));
+
+  playButton.addEventListener("click", togglePlay);
+  seek.addEventListener("input", () => { const position = Number(seek.value); if (Number.isFinite(position)) { audio.currentTime = position; updateText(); draw(); } });
+  speed.addEventListener("change", () => { const rate = Number(speed.value); if (Number.isFinite(rate) && rate > 0) audio.playbackRate = rate; });
+  if (typeof ResizeObserver !== "undefined") { resizeObserver = new ResizeObserver(draw); resizeObserver.observe(canvas); }
+
+  function setSamples(value) {
+    if (!(value instanceof Float32Array)) throw new TypeError("波形データはFloat32Arrayで指定してください。");
+    samples = value.slice(); draw();
+  }
+  function dispose() {
+    disposed = true; stopAnimation(); resizeObserver?.disconnect(); resizeObserver = null;
+    listeners.forEach(([event, handler]) => audio.removeEventListener(event, handler));
+    playButton.removeEventListener("click", togglePlay);
+    audio.pause();
+  }
+
+  const initialRate = Number(speed.value); if (Number.isFinite(initialRate) && initialRate > 0) audio.playbackRate = initialRate;
+  setButton(false); updateText(); setState("idle"); draw();
+  return { setSamples, refreshTheme: draw, dispose };
+}

audio-recorder.js

@@ -0,0 +1,167 @@
+const MAX_SECONDS = 30;
+
+const microphoneIcon = `<svg aria-hidden="true" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round"><rect x="8" y="3" width="8" height="12" rx="4"/><path d="M5 11a7 7 0 0 0 14 0M12 18v3M8 21h8"/></svg>`;
+const stopIcon = `<svg aria-hidden="true" viewBox="0 0 24 24" fill="currentColor"><rect x="7" y="7" width="10" height="10" rx="1"/></svg>`;
+
+function formatTime(seconds) {
+  const whole = Math.min(MAX_SECONDS, Math.max(0, Math.floor(seconds)));
+  return `${String(Math.floor(whole / 60)).padStart(2, "0")}:${String(whole % 60).padStart(2, "0")}`;
+}
+
+function messageFor(error) {
+  if (error?.name === "NotAllowedError" || error?.name === "SecurityError") {
+    return "マイクの使用が許可されていません。ブラウザの設定でマイクを許可してから、もう一度お試しください。";
+  }
+  if (error?.name === "NotFoundError") return "利用できるマイクが見つかりません。マイクを接続してから、もう一度お試しください。";
+  if (error?.name === "NotReadableError") return "マイクを使用できません。他のアプリで使用中でないか確認してください。";
+  if (error?.name === "NotSupportedError") return "このブラウザでは録音を利用できません。最新版の対応ブラウザでお試しください。";
+  return "録音を開始できませんでした。ブラウザとマイクの状態を確認して、もう一度お試しください。";
+}
+
+/**
+ * Creates a local-only microphone recorder. `onAudio` receives a finished Blob.
+ */
+export function createAudioRecorder({ button, canvas, timer, state, onAudio, onError } = {}) {
+  if (!button || !canvas || !timer || !state) throw new Error("録音UIの要素が不足しています。");
+
+  const recorderContainer = button.closest(".recorder");
+  const context2d = canvas.getContext("2d");
+  let phase = "idle";
+  let mediaRecorder = null;
+  let stream = null;
+  let audioContext = null;
+  let analyser = null;
+  let animationFrame = 0;
+  let lastDrawAt = 0;
+  let startedAt = 0;
+  let deadlineTimer = 0;
+  let ticker = 0;
+  let chunks = [];
+
+  canvas.setAttribute("role", "img");
+  canvas.setAttribute("aria-label", "録音中の音声波形");
+
+  function setState(text, kind = "") {
+    state.textContent = text;
+    state.dataset.state = kind;
+  }
+
+  function setButton(recording = false, disabled = false) {
+    button.disabled = disabled;
+    button.setAttribute("aria-pressed", String(recording));
+    button.innerHTML = recording ? `${stopIcon}<span>録音を終了</span>` : `${microphoneIcon}<span>録音をはじめる</span>`;
+    recorderContainer?.classList.toggle("is-recording", recording);
+  }
+
+  function drawIdle() {
+    const rect = canvas.getBoundingClientRect();
+    const dpr = window.devicePixelRatio || 1;
+    const width = Math.max(1, Math.round(rect.width * dpr));
+    const height = Math.max(1, Math.round(rect.height * dpr));
+    if (canvas.width !== width || canvas.height !== height) { canvas.width = width; canvas.height = height; }
+    context2d.clearRect(0, 0, width, height);
+    context2d.strokeStyle = "rgba(120, 134, 164, .38)";
+    context2d.lineWidth = Math.max(1, dpr);
+    context2d.beginPath(); context2d.moveTo(0, height / 2); context2d.lineTo(width, height / 2); context2d.stroke();
+  }
+
+  function drawWave(now) {
+    animationFrame = requestAnimationFrame(drawWave);
+    if (!analyser || now - lastDrawAt < 50) return;
+    lastDrawAt = now;
+    const rect = canvas.getBoundingClientRect();
+    const dpr = window.devicePixelRatio || 1;
+    const width = Math.max(1, Math.round(rect.width * dpr));
+    const height = Math.max(1, Math.round(rect.height * dpr));
+    if (canvas.width !== width || canvas.height !== height) { canvas.width = width; canvas.height = height; }
+    const samples = new Uint8Array(analyser.fftSize);
+    analyser.getByteTimeDomainData(samples);
+    context2d.clearRect(0, 0, width, height);
+    context2d.strokeStyle = "#e34a52";
+    context2d.lineWidth = Math.max(1.5, 2 * dpr);
+    context2d.lineJoin = "round";
+    context2d.beginPath();
+    for (let index = 0; index < samples.length; index += 1) {
+      const x = index / (samples.length - 1) * width;
+      const y = (samples[index] / 255) * height;
+      if (index === 0) context2d.moveTo(x, y); else context2d.lineTo(x, y);
+    }
+    context2d.stroke();
+  }
+
+  function updateTimer() {
+    timer.textContent = `${formatTime((performance.now() - startedAt) / 1000)} / 0:30`;
+  }
+
+  function stopVisuals() {
+    cancelAnimationFrame(animationFrame); animationFrame = 0;
+    clearInterval(ticker); ticker = 0;
+    clearTimeout(deadlineTimer); deadlineTimer = 0;
+    drawIdle();
+  }
+
+  async function closeAudio() {
+    if (audioContext && audioContext.state !== "closed") await audioContext.close().catch(() => {});
+    audioContext = null; analyser = null;
+    stream?.getTracks().forEach((track) => track.stop()); stream = null;
+  }
+
+  async function finishRecording() {
+    const blob = new Blob(chunks, { type: mediaRecorder?.mimeType || "audio/webm" });
+    mediaRecorder = null; chunks = [];
+    stopVisuals(); await closeAudio();
+    phase = "processing"; setButton(false, true); setState("録音を読み込んでいます…", "processing");
+    try {
+      await onAudio?.(blob);
+      phase = "idle"; setButton(); setState("録音を読み込みました。再生して内容を確認できます。", "recorded");
+    } catch (error) {
+      phase = "idle"; setButton(); setState("録音を読み込めませんでした。もう一度お試しください。", "error");
+      onError?.(error);
+    }
+  }
+
+  function stopRecording() {
+    if (phase !== "recording") return;
+    phase = "stopping"; setButton(false, true); setState("録音を終了しています…", "processing");
+    stopVisuals();
+    if (mediaRecorder?.state === "recording") mediaRecorder.stop();
+  }
+
+  async function startRecording() {
+    if (phase !== "idle") return;
+    if (!navigator.mediaDevices?.getUserMedia || !window.MediaRecorder) {
+      const error = new DOMException("MediaRecorder is unavailable", "NotSupportedError");
+      setState(messageFor(error), "error"); onError?.(error); return;
+    }
+    phase = "requesting"; setButton(false, true); setState("マイクの許可を待っています…", "pending");
+    try {
+      stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+      audioContext = new AudioContext();
+      analyser = audioContext.createAnalyser(); analyser.fftSize = 1024;
+      const source = audioContext.createMediaStreamSource(stream); source.connect(analyser);
+      mediaRecorder = new MediaRecorder(stream);
+      chunks = [];
+      mediaRecorder.ondataavailable = (event) => { if (event.data.size) chunks.push(event.data); };
+      mediaRecorder.onstop = () => { finishRecording(); };
+      mediaRecorder.onerror = (event) => { const error = event.error || new Error("録音中に問題が発生しました。"); setState("録音中に問題が発生しました。もう一度お試しください。", "error"); onError?.(error); };
+      startedAt = performance.now(); phase = "recording";
+      mediaRecorder.start(); setButton(true); setState("", "recording"); updateTimer();
+      ticker = window.setInterval(updateTimer, 250);
+      deadlineTimer = window.setTimeout(stopRecording, MAX_SECONDS * 1000);
+      lastDrawAt = 0; animationFrame = requestAnimationFrame(drawWave);
+    } catch (error) {
+      await closeAudio(); stopVisuals(); phase = "idle"; setButton(); setState(messageFor(error), "error"); onError?.(error);
+    }
+  }
+
+  function toggle() { if (phase === "recording") stopRecording(); else if (phase === "idle") startRecording(); }
+  function dispose() {
+    button.removeEventListener("click", toggle);
+    if (mediaRecorder) { mediaRecorder.onstop = null; mediaRecorder.onerror = null; if (mediaRecorder.state === "recording") mediaRecorder.stop(); }
+    stopVisuals(); closeAudio(); phase = "disposed";
+  }
+
+  button.addEventListener("click", toggle);
+  setButton(); timer.textContent = "00:00 / 0:30"; setState("録音の準備ができました。", "idle"); drawIdle();
+  return { dispose, stop: stopRecording };
+}

examples/ADULT_VOICE_PROVENANCE.json

@@ -0,0 +1,89 @@
+[
+  {
+    "file": "adult-reservation-v3.wav",
+    "kind": "reservation",
+    "text": "もしもし、あしたの夜七時に、二人で予約できますか。",
+    "caption": "40代の成人男性の中低音の声。飾らない日常会話。抑揚は小さく、落ち着いた普通の話し方。語尾を伸ばさず、力を抜いて自然に話す。",
+    "seed": 5712,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "reference_audio": null,
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.56,
+    "max_seconds": 8.0,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "generation_seconds": 0.5680867279879749,
+    "peak": 0.9645974040031433,
+    "rms": 0.15562847256660461,
+    "sha256": "e494ae99d12c8fd23ba4741a5a8507942f0e54877b1c43493d5924e0a03c7585",
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "review_note": "Caption specifies voice style. No perceptual listening claim.",
+    "source_file": "reservation-5712.wav",
+    "creator": "mocomoco inc.; synthesized with Irodori VoiceDesign by Chihiro Arata"
+  },
+  {
+    "file": "adult-backchannel-v3.wav",
+    "kind": "backchannel",
+    "text": "ああ、そうなんですね。",
+    "caption": "40代の成人男性の中低音の声。飾らない日常会話。抑揚は小さく、落ち着いた普通の話し方。語尾を伸ばさず、力を抜いて自然に話す。",
+    "seed": 5701,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "reference_audio": null,
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.48,
+    "max_seconds": 3.5,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "generation_seconds": 0.9337333729490638,
+    "peak": 0.8861069679260254,
+    "rms": 0.12877875566482544,
+    "sha256": "424c3152da9a8b1444363ff71574cfe26c855b5fe65e22ab6a553530132bdbe5",
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "review_note": "Caption specifies voice style. No perceptual listening claim.",
+    "source_file": "backchannel-5701.wav",
+    "creator": "mocomoco inc.; synthesized with Irodori VoiceDesign by Chihiro Arata"
+  },
+  {
+    "file": "adult-cancel-v3.wav",
+    "kind": "cancel",
+    "text": "すみません、あしたの予約をキャンセルしたいんですけど。",
+    "caption": "40代の成人男性の中低音の声。飾らない日常会話。抑揚は小さく、落ち着いた普通の話し方。語尾を伸ばさず、力を抜いて自然に話す。",
+    "seed": 5721,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "reference_audio": null,
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.92,
+    "max_seconds": 8.0,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "generation_seconds": 0.511951609980315,
+    "peak": 0.9111555218696594,
+    "rms": 0.14880435168743134,
+    "sha256": "27a758e3f6db7c2553db937a638f1ddf60e9ad79d3c5f49a8d8097480005dd99",
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "review_note": "Caption specifies voice style. No perceptual listening claim.",
+    "source_file": "cancel-5721.wav",
+    "creator": "mocomoco inc.; synthesized with Irodori VoiceDesign by Chihiro Arata"
+  }
+]

examples/ATTRIBUTION.json

@@ -0,0 +1,599 @@
+[
+  {
+    "file": "reservation.wav",
+    "description": "Original authored Japanese text synthesized using Mei voice",
+    "voice": "mei_normal.htsvoice",
+    "creator": "Nagoya Institute of Technology / MMDAgent project",
+    "license": "CC-BY-3.0",
+    "source": "https://www.mmdagent.jp/",
+    "changes": "Synthesized speech, mono 16kHz WAV"
+  },
+  {
+    "file": "music-639.wav",
+    "license": "CC0-1.0",
+    "creator": "memexikon",
+    "license_url": "http://creativecommons.org/publicdomain/zero/1.0/",
+    "source_url": "https://freesound.org/people/memexikon/sounds/639/",
+    "title": "lockjawnetf.wav",
+    "changes": "None; original source WAV"
+  },
+  {
+    "file": "animal-dog-birds-238688.wav",
+    "creator": "SoundEnsemble",
+    "title": "SE ANIMALS dog barking birds.wav",
+    "source_url": "https://freesound.org/people/SoundEnsemble/sounds/238688/",
+    "preview_url": "https://cdn.freesound.org/previews/238/238688_3662372-hq.mp3",
+    "license": "CC0-1.0",
+    "license_url": "https://creativecommons.org/publicdomain/zero/1.0/",
+    "license_verified": "2026-09-20",
+    "changes": "Freesound high-quality preview decoded, downmixed to mono, and resampled to 16 kHz PCM WAV; full 22.248688-second recording retained.",
+    "source_sha256": "c2e6ea3ebff38bb72c57831da11e22fb606b845e9bd6085155e43dc64be788f9",
+    "sha256": "cb229bbe56c08902fb2e1e1d7bad639b49908148e384bdf3a1aa8d9687b19c24",
+    "source_id_audit": "freesound:238688 was absent from local FSD50K and typed-audio manifests at selection time.",
+    "selection_note": "Selected from its dog-and-birds description and CC0 license before inference; no model output was used to choose it. Exclude this demo from future independent evaluation."
+  },
+  {
+    "file": "irodori-reservation.wav",
+    "text": "明日の夜七時に、二人で予約をお願いできますか。",
+    "expected_intent": "reservation",
+    "synthetic_audio": true,
+    "model": "Aratako/Irodori-TTS-500M-v3",
+    "revision": "236c1e56591279fc24e3c1bf6609fc06e48dde28",
+    "source_code_revision": "eaf74d6a19138f743acb5b71a445fd25a57db987",
+    "seed": 913,
+    "num_steps": 40,
+    "reference_audio": null,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.0,
+    "generation_seconds": 0.9263423918746412,
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-500M-v3",
+    "creator": "MocoVoice; synthesized with Irodori-TTS by Chihiro Arata",
+    "sha256": "8f8ee91977655e119b26d3d7d35e9b414e62f1ef9c6b2c46097f010442f61fd9"
+  },
+  {
+    "file": "irodori-change.wav",
+    "text": "すみません、予約を土曜日の午後三時に変更したいのですが。",
+    "expected_intent": "change",
+    "synthetic_audio": true,
+    "model": "Aratako/Irodori-TTS-500M-v3",
+    "revision": "236c1e56591279fc24e3c1bf6609fc06e48dde28",
+    "source_code_revision": "eaf74d6a19138f743acb5b71a445fd25a57db987",
+    "seed": 271,
+    "num_steps": 40,
+    "reference_audio": null,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.72,
+    "generation_seconds": 0.4987814510241151,
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-500M-v3",
+    "creator": "MocoVoice; synthesized with Irodori-TTS by Chihiro Arata",
+    "sha256": "5c3fe27ad38a581e1f545c6cc8b18f8dd7a75b24e7c2737952e3b0e88ae589c2"
+  },
+  {
+    "file": "irodori-cancel.wav",
+    "text": "急な予定が入ってしまったので、明日の予約をキャンセルしてください。",
+    "expected_intent": "cancel",
+    "synthetic_audio": true,
+    "model": "Aratako/Irodori-TTS-500M-v3",
+    "revision": "236c1e56591279fc24e3c1bf6609fc06e48dde28",
+    "source_code_revision": "eaf74d6a19138f743acb5b71a445fd25a57db987",
+    "seed": 614,
+    "num_steps": 40,
+    "reference_audio": null,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.88,
+    "generation_seconds": 0.49401673721149564,
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-500M-v3",
+    "creator": "MocoVoice; synthesized with Irodori-TTS by Chihiro Arata",
+    "sha256": "25945f9ea6923cb47a1b008e0dbe935e4ee30bb3a30a1a030c12817b6a693b01"
+  },
+  {
+    "file": "natural-reservation-a.wav",
+    "intent": "reservation",
+    "variant": "a",
+    "text": "こんばんは。あしたの夜、しちじごろに二人でお願いしたいんですけど、まだ空いてますか？",
+    "caption": "落ち着いた大人の話者。予約の電話で、相手に配慮しながら、自然な会話の速さでやわらかく話す。文の切れ目では短く息をつぎ、丁寧だが読み上げ調にならない。",
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "model_card_license_verified": "2026-09-20",
+    "source_code_revision": "unversioned-worktree",
+    "seed": 1081,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 8.44,
+    "peak_abs": 0.9765087366104126,
+    "rms": 0.1376929726771456,
+    "generation_seconds": 1.1542682349681854,
+    "sha256": "bf9ce6d83ae7fe8961f97f4cd95a7d6be1ca29d2c373cc77c96b7a41d7d719d2",
+    "integrity_note": "Duration, peak, RMS, and SHA-256 are file checks only; they do not measure naturalness.",
+    "provenance_note": "Text-plus-caption-only synthesis; no real-person reference audio or voice embedding supplied."
+  },
+  {
+    "file": "natural-change-a.wav",
+    "intent": "change",
+    "variant": "a",
+    "text": "すみません、予約のことなんですけど。土曜日の午後、さんじに変更できますか？",
+    "caption": "穏やかで親しみのある大人の話者。予約内容の変更を電話で相談している。急がず自然な間を取り、控えめにお願いする会話らしい口調。",
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "model_card_license_verified": "2026-09-20",
+    "source_code_revision": "unversioned-worktree",
+    "seed": 2081,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 6.64,
+    "peak_abs": 0.7287325859069824,
+    "rms": 0.1640570232803877,
+    "generation_seconds": 0.627793378662318,
+    "sha256": "a1a9828f1beada7c90182df61d30872a5edc056bef833219fd53fc48ee243a2d",
+    "integrity_note": "Duration, peak, RMS, and SHA-256 are file checks only; they do not measure naturalness.",
+    "provenance_note": "Text-plus-caption-only synthesis; no real-person reference audio or voice embedding supplied."
+  },
+  {
+    "file": "natural-cancel-b.wav",
+    "intent": "cancel",
+    "variant": "b",
+    "text": "すみません、急に予定が入ってしまって…。あしたの予約、キャンセルをお願いできますか？",
+    "caption": "落ち着いた大人の話者。予定変更を申し訳なく思いながら、予約のキャンセルを電話で伝える。小さなためらいと自然な息継ぎがあり、丁寧でやわらかい口調。",
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "model_card_license_verified": "2026-09-20",
+    "source_code_revision": "unversioned-worktree",
+    "seed": 3082,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 8.96,
+    "peak_abs": 0.8194587230682373,
+    "rms": 0.1417061149254557,
+    "generation_seconds": 0.7651458447799087,
+    "sha256": "52f0ed8dfd7ef2cbdd0ce67533f4f1acf2fe7ced93c9e34aa5d49957609f35dc",
+    "integrity_note": "Duration, peak, RMS, and SHA-256 are file checks only; they do not measure naturalness.",
+    "provenance_note": "Text-plus-caption-only synthesis; no real-person reference audio or voice embedding supplied."
+  },
+  {
+    "id": "order-full",
+    "name": "注文を最後まで伝える",
+    "speech_act": "request",
+    "intent": "order",
+    "turn_complete": true,
+    "context": "店員: ご注文が決まりましたら、お声がけください。",
+    "text": "きのこのパスタを一つ、持ち帰りでお願いします。",
+    "caption": "落ち着いた大人の話者。店員に注文を伝える短い会話。自然な速さで、最後まで言い切る。",
+    "seed": 4101,
+    "file": "turn-order-full.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "order",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.08,
+    "peak_abs": 0.7851645946502686,
+    "rms": 0.17343387180324984,
+    "generation_seconds": 0.9514303421601653,
+    "sha256": "1cd199edb025dc5e93bb8ef22dc8cb98297ea879ee30e578bef43cd2c0260618",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "backchannel-full",
+    "name": "相手の説明への相づち",
+    "speech_act": "backchannel",
+    "intent": "none",
+    "turn_complete": true,
+    "context": "案内係: 入口を出たら右へ進むと、受付が見えてきます。",
+    "text": "うん、分かりました。",
+    "caption": "親しみのある大人の話者。相手の説明を聞きながら短く相づちを打つ。急がず、自然に言い切る。",
+    "seed": 4102,
+    "file": "turn-backchannel-full.wav",
+    "expected": {
+      "speech_act": "backchannel",
+      "intent": "none",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.24,
+    "peak_abs": 0.7609981894493103,
+    "rms": 0.15303109302071377,
+    "generation_seconds": 0.544307358097285,
+    "sha256": "ab9c4c35fa15521b227540689743c15db1fae9a4701cd1c22069fe0789ac844d",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "confirm-full",
+    "name": "内容を確認して承認する",
+    "speech_act": "confirm",
+    "intent": "reservation",
+    "turn_complete": true,
+    "context": "店員: 来週の火曜日、午後六時に二名様でお席をお取りしてよろしいですか。",
+    "text": "はい、その内容でお願いします。",
+    "caption": "穏やかな大人の話者。店員の確認に簡潔に同意する。会話らしい自然な間で、やわらかく言い切る。",
+    "seed": 4103,
+    "file": "turn-confirm-full.wav",
+    "expected": {
+      "speech_act": "confirm",
+      "intent": "reservation",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.16,
+    "peak_abs": 0.6248125433921814,
+    "rms": 0.15614677213260153,
+    "generation_seconds": 0.49432186922058463,
+    "sha256": "5c9569279a31be36d519b3e9e924e187d29e3eba967ed70850915108e4590e51",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "correction-full",
+    "name": "日付を訂正して変更する",
+    "speech_act": "correct",
+    "intent": "change",
+    "turn_complete": true,
+    "context": "店員: ご予約は金曜日の午後三時で承っています。",
+    "text": "あ、すみません。金曜日ではなく、土曜日の午後に変更をお願いします。",
+    "caption": "丁寧な大人の話者。予約内容を思い出して訂正する。最初に短いためらいを置き、最後ははっきり言い切る。",
+    "seed": 4104,
+    "file": "turn-correction-full.wav",
+    "expected": {
+      "speech_act": "correct",
+      "intent": "change",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 6.4,
+    "peak_abs": 0.7035095691680908,
+    "rms": 0.1609569216640416,
+    "generation_seconds": 0.6292441417463124,
+    "sha256": "b73eb1847645cd473f71a8d8296639f6bc38ec17dc628ec843a680d005d44a1c",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "cancel-full",
+    "name": "注文を取り消す",
+    "speech_act": "cancel",
+    "intent": "cancel",
+    "turn_complete": true,
+    "context": "店員: 先ほどのスープとコーヒーのご注文を確認します。",
+    "text": "すみません、スープのほうは取り消してください。",
+    "caption": "落ち着いた大人の話者。注文の一部を丁寧に取り消す。自然な会話の速さで、申し訳なさを少し含めて言い切る。",
+    "seed": 4105,
+    "file": "turn-cancel-full.wav",
+    "expected": {
+      "speech_act": "cancel",
+      "intent": "cancel",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.44,
+    "peak_abs": 0.781799852848053,
+    "rms": 0.16470869580552724,
+    "generation_seconds": 0.5071098008193076,
+    "sha256": "24ca56ecc4191620c37834cf5d2f28c51ed7192ee5b6235585fb2dcc0ed38045",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "order-hesitation-unfinished",
+    "name": "迷って途中で止まる注文",
+    "speech_act": "request",
+    "intent": "order",
+    "turn_complete": false,
+    "context": "店員: ご注文が決まりましたら、お声がけください。",
+    "text": "きのこのパスタを一つ、ええと、持ち帰りにするか…",
+    "caption": "迷っている大人の話者。注文を考えながら言いよどみ、答えを決めずに途中で止める。読み上げ調にしない。",
+    "seed": 4106,
+    "file": "turn-order-hesitation-unfinished.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "order",
+      "turn_complete": false
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 5.08,
+    "peak_abs": 0.9923233389854431,
+    "rms": 0.161138963203946,
+    "generation_seconds": 0.5771061358973384,
+    "sha256": "70c4bc0af49280d772c09ea9977f5a1d072f58c71705be9dc36bd1704010ea1e",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "reservation-clause-unfinished",
+    "name": "意味のある途中節で止まる予約",
+    "speech_act": "request",
+    "intent": "reservation",
+    "turn_complete": false,
+    "context": "店員: ご希望の日時と人数をお知らせください。",
+    "text": "あしたの夕方に二人で、もし窓際が…",
+    "caption": "控えめな大人の話者。予約の条件を伝え始めるが、文を完結させずに意味のある途中で止まる。自然な短い間を含める。",
+    "seed": 4107,
+    "file": "turn-reservation-clause-unfinished.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "reservation",
+      "turn_complete": false
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 5.28,
+    "peak_abs": 1.0007697343826294,
+    "rms": 0.13793335127204556,
+    "generation_seconds": 0.5772101040929556,
+    "sha256": "14de328d4e6a4271e671967333fc0c5392d637d0a9876c7b6d0d72a1eb4fa926",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "reservation-full-contrast",
+    "name": "同じ予約条件を最後まで伝える",
+    "speech_act": "request",
+    "intent": "reservation",
+    "turn_complete": true,
+    "context": "店員: ご希望の日時と人数をお知らせください。",
+    "text": "あしたの夕方に二人で、窓際の席を予約したいです。",
+    "caption": "控えめな大人の話者。予約の条件を伝え、最後まで自然に言い切る。文の区切りに短い息継ぎを入れる。",
+    "seed": 4108,
+    "file": "turn-reservation-full-contrast.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "reservation",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 5.68,
+    "peak_abs": 0.8254440426826477,
+    "rms": 0.15123665222928692,
+    "generation_seconds": 0.5876368628814816,
+    "sha256": "36a84d9c7061577b438529f4e0882d456f7f3ce773cc9f55248be5fcac1a43a3",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "file": "adult-reservation-v3.wav",
+    "kind": "reservation",
+    "text": "もしもし、あしたの夜七時に、二人で予約できますか。",
+    "caption": "40代の成人男性の中低音の声。飾らない日常会話。抑揚は小さく、落ち着いた普通の話し方。語尾を伸ばさず、力を抜いて自然に話す。",
+    "seed": 5712,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "reference_audio": null,
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.56,
+    "max_seconds": 8.0,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "generation_seconds": 0.5680867279879749,
+    "peak": 0.9645974040031433,
+    "rms": 0.15562847256660461,
+    "sha256": "e494ae99d12c8fd23ba4741a5a8507942f0e54877b1c43493d5924e0a03c7585",
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "review_note": "Caption specifies voice style. No perceptual listening claim.",
+    "source_file": "reservation-5712.wav",
+    "creator": "mocomoco inc.; synthesized with Irodori VoiceDesign by Chihiro Arata"
+  },
+  {
+    "file": "adult-backchannel-v3.wav",
+    "kind": "backchannel",
+    "text": "ああ、そうなんですね。",
+    "caption": "40代の成人男性の中低音の声。飾らない日常会話。抑揚は小さく、落ち着いた普通の話し方。語尾を伸ばさず、力を抜いて自然に話す。",
+    "seed": 5701,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "reference_audio": null,
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.48,
+    "max_seconds": 3.5,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "generation_seconds": 0.9337333729490638,
+    "peak": 0.8861069679260254,
+    "rms": 0.12877875566482544,
+    "sha256": "424c3152da9a8b1444363ff71574cfe26c855b5fe65e22ab6a553530132bdbe5",
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "review_note": "Caption specifies voice style. No perceptual listening claim.",
+    "source_file": "backchannel-5701.wav",
+    "creator": "mocomoco inc.; synthesized with Irodori VoiceDesign by Chihiro Arata"
+  },
+  {
+    "file": "adult-cancel-v3.wav",
+    "kind": "cancel",
+    "text": "すみません、あしたの予約をキャンセルしたいんですけど。",
+    "caption": "40代の成人男性の中低音の声。飾らない日常会話。抑揚は小さく、落ち着いた普通の話し方。語尾を伸ばさず、力を抜いて自然に話す。",
+    "seed": 5721,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "reference_audio": null,
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.92,
+    "max_seconds": 8.0,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "generation_seconds": 0.511951609980315,
+    "peak": 0.9111555218696594,
+    "rms": 0.14880435168743134,
+    "sha256": "27a758e3f6db7c2553db937a638f1ddf60e9ad79d3c5f49a8d8097480005dd99",
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "review_note": "Caption specifies voice style. No perceptual listening claim.",
+    "source_file": "cancel-5721.wav",
+    "creator": "mocomoco inc.; synthesized with Irodori VoiceDesign by Chihiro Arata"
+  },
+  {
+    "file": "music-acoustic-loop-848704.wav",
+    "creator": "EEE3333E",
+    "title": "Acoustic Guitar loop",
+    "source_url": "https://freesound.org/people/EEE3333E/sounds/848704/",
+    "preview_url": "https://cdn.freesound.org/previews/848/848704_17604172-hq.mp3",
+    "license": "CC0-1.0",
+    "license_url": "https://creativecommons.org/publicdomain/zero/1.0/",
+    "license_verified": "2026-09-20",
+    "changes": "Freesound high-quality preview decoded, downmixed to mono, and resampled to 16 kHz PCM WAV; full 16.045-second loop retained.",
+    "source_sha256": "93a4c8a5c06977e94b22a41a280aa3360bd7f1f61ecf68cb5cc66e76f0ffa390",
+    "sha256": "c851b6cad67ac9ef08f5c634efb90f001670f5ed454320c276310bfe9b7c020a",
+    "source_id_audit": "freesound:848704 was absent from local FSD50K and typed-audio manifests at selection time.",
+    "selection_note": "Selected from its acoustic-guitar-loop description and CC0 license before inference; no model output was used to choose it. Exclude this demo from future independent evaluation."
+  }
+]

examples/IRODORI_PROVENANCE.json

@@ -0,0 +1,62 @@
+[
+  {
+    "file": "irodori-reservation.wav",
+    "text": "明日の夜七時に、二人で予約をお願いできますか。",
+    "expected_intent": "reservation",
+    "synthetic_audio": true,
+    "model": "Aratako/Irodori-TTS-500M-v3",
+    "revision": "236c1e56591279fc24e3c1bf6609fc06e48dde28",
+    "source_code_revision": "eaf74d6a19138f743acb5b71a445fd25a57db987",
+    "seed": 913,
+    "num_steps": 40,
+    "reference_audio": null,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.0,
+    "generation_seconds": 0.9263423918746412,
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-500M-v3",
+    "creator": "MocoVoice; synthesized with Irodori-TTS by Chihiro Arata",
+    "sha256": "8f8ee91977655e119b26d3d7d35e9b414e62f1ef9c6b2c46097f010442f61fd9"
+  },
+  {
+    "file": "irodori-change.wav",
+    "text": "すみません、予約を土曜日の午後三時に変更したいのですが。",
+    "expected_intent": "change",
+    "synthetic_audio": true,
+    "model": "Aratako/Irodori-TTS-500M-v3",
+    "revision": "236c1e56591279fc24e3c1bf6609fc06e48dde28",
+    "source_code_revision": "eaf74d6a19138f743acb5b71a445fd25a57db987",
+    "seed": 271,
+    "num_steps": 40,
+    "reference_audio": null,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.72,
+    "generation_seconds": 0.4987814510241151,
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-500M-v3",
+    "creator": "MocoVoice; synthesized with Irodori-TTS by Chihiro Arata",
+    "sha256": "5c3fe27ad38a581e1f545c6cc8b18f8dd7a75b24e7c2737952e3b0e88ae589c2"
+  },
+  {
+    "file": "irodori-cancel.wav",
+    "text": "急な予定が入ってしまったので、明日の予約をキャンセルしてください。",
+    "expected_intent": "cancel",
+    "synthetic_audio": true,
+    "model": "Aratako/Irodori-TTS-500M-v3",
+    "revision": "236c1e56591279fc24e3c1bf6609fc06e48dde28",
+    "source_code_revision": "eaf74d6a19138f743acb5b71a445fd25a57db987",
+    "seed": 614,
+    "num_steps": 40,
+    "reference_audio": null,
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.88,
+    "generation_seconds": 0.49401673721149564,
+    "license": "MIT",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-500M-v3",
+    "creator": "MocoVoice; synthesized with Irodori-TTS by Chihiro Arata",
+    "sha256": "25945f9ea6923cb47a1b008e0dbe935e4ee30bb3a30a1a030c12817b6a693b01"
+  }
+]

examples/PHONE_PROVENANCE.json

@@ -0,0 +1,42 @@
+[
+  {
+    "file": "natural-reservation-a-telephone.wav",
+    "source_file": "natural-reservation-a.wav",
+    "source_sha256": "bf9ce6d83ae7fe8961f97f4cd95a7d6be1ca29d2c373cc77c96b7a41d7d719d2",
+    "sha256": "9c30790ee4ef54eb4e12f7ea58ff1cd2d65d399560f174e61d5a2451b0aa7f77",
+    "transformation": "300–3400 Hz bandpass, 8 kHz mono, 8-bit mu-law companding approximation decoded to PCM16",
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_note": "Original synthesis watermarked; retention after telephone processing has not been verified."
+  },
+  {
+    "file": "turn-reservation-clause-unfinished-telephone.wav",
+    "source_file": "turn-reservation-clause-unfinished.wav",
+    "source_sha256": "14de328d4e6a4271e671967333fc0c5392d637d0a9876c7b6d0d72a1eb4fa926",
+    "sha256": "ea3a9e6b5bb09ab2cbf6263a92d46bb91baabc18ff1094e3cc65de24af20a499",
+    "transformation": "300–3400 Hz bandpass, 8 kHz mono, 8-bit mu-law companding approximation decoded to PCM16",
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_note": "Original synthesis watermarked; retention after telephone processing has not been verified."
+  },
+  {
+    "file": "adult-reservation-v3-telephone.wav",
+    "source_file": "adult-reservation-v3.wav",
+    "source_sha256": "e494ae99d12c8fd23ba4741a5a8507942f0e54877b1c43493d5924e0a03c7585",
+    "sha256": "a281dac7872dfca1b2ca0ea38de4c7ed46af51f32e33ef2091ee0cb0efa59c52",
+    "transformation": "300–3400 Hz bandpass, 8 kHz mono, 8-bit mu-law companding approximation decoded to PCM16",
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_note": "Original synthesis watermarked; retention after telephone processing has not been verified."
+  },
+  {
+    "file": "adult-cancel-v3-telephone.wav",
+    "source_file": "adult-cancel-v3.wav",
+    "source_sha256": "27a758e3f6db7c2553db937a638f1ddf60e9ad79d3c5f49a8d8097480005dd99",
+    "sha256": "7657c60e23fce2e30998d072908d726163ae33c84cb8c6008248c0a27a4c7087",
+    "transformation": "300–3400 Hz bandpass, 8 kHz mono, 8-bit mu-law companding approximation decoded to PCM16",
+    "synthetic_audio": true,
+    "training_excluded": true,
+    "watermark_note": "Original synthesis watermarked; retention after telephone processing has not been verified."
+  }
+]

examples/VOICE_DESIGN_PROVENANCE.json

@@ -0,0 +1,404 @@
+[
+  {
+    "file": "natural-reservation-a.wav",
+    "intent": "reservation",
+    "variant": "a",
+    "text": "こんばんは。あしたの夜、しちじごろに二人でお願いしたいんですけど、まだ空いてますか？",
+    "caption": "落ち着いた大人の話者。予約の電話で、相手に配慮しながら、自然な会話の速さでやわらかく話す。文の切れ目では短く息をつぎ、丁寧だが読み上げ調にならない。",
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "model_card_license_verified": "2026-09-20",
+    "source_code_revision": "unversioned-worktree",
+    "seed": 1081,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 8.44,
+    "peak_abs": 0.9765087366104126,
+    "rms": 0.1376929726771456,
+    "generation_seconds": 1.1542682349681854,
+    "sha256": "bf9ce6d83ae7fe8961f97f4cd95a7d6be1ca29d2c373cc77c96b7a41d7d719d2",
+    "integrity_note": "Duration, peak, RMS, and SHA-256 are file checks only; they do not measure naturalness.",
+    "provenance_note": "Text-plus-caption-only synthesis; no real-person reference audio or voice embedding supplied."
+  },
+  {
+    "file": "natural-change-a.wav",
+    "intent": "change",
+    "variant": "a",
+    "text": "すみません、予約のことなんですけど。土曜日の午後、さんじに変更できますか？",
+    "caption": "穏やかで親しみのある大人の話者。予約内容の変更を電話で相談している。急がず自然な間を取り、控えめにお願いする会話らしい口調。",
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "model_card_license_verified": "2026-09-20",
+    "source_code_revision": "unversioned-worktree",
+    "seed": 2081,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 6.64,
+    "peak_abs": 0.7287325859069824,
+    "rms": 0.1640570232803877,
+    "generation_seconds": 0.627793378662318,
+    "sha256": "a1a9828f1beada7c90182df61d30872a5edc056bef833219fd53fc48ee243a2d",
+    "integrity_note": "Duration, peak, RMS, and SHA-256 are file checks only; they do not measure naturalness.",
+    "provenance_note": "Text-plus-caption-only synthesis; no real-person reference audio or voice embedding supplied."
+  },
+  {
+    "file": "natural-cancel-b.wav",
+    "intent": "cancel",
+    "variant": "b",
+    "text": "すみません、急に予定が入ってしまって…。あしたの予約、キャンセルをお願いできますか？",
+    "caption": "落ち着いた大人の話者。予定変更を申し訳なく思いながら、予約のキャンセルを電話で伝える。小さなためらいと自然な息継ぎがあり、丁寧でやわらかい口調。",
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "model_card_license_verified": "2026-09-20",
+    "source_code_revision": "unversioned-worktree",
+    "seed": 3082,
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 8.96,
+    "peak_abs": 0.8194587230682373,
+    "rms": 0.1417061149254557,
+    "generation_seconds": 0.7651458447799087,
+    "sha256": "52f0ed8dfd7ef2cbdd0ce67533f4f1acf2fe7ced93c9e34aa5d49957609f35dc",
+    "integrity_note": "Duration, peak, RMS, and SHA-256 are file checks only; they do not measure naturalness.",
+    "provenance_note": "Text-plus-caption-only synthesis; no real-person reference audio or voice embedding supplied."
+  },
+  {
+    "id": "order-full",
+    "name": "注文を最後まで伝える",
+    "speech_act": "request",
+    "intent": "order",
+    "turn_complete": true,
+    "context": "店員: ご注文が決まりましたら、お声がけください。",
+    "text": "きのこのパスタを一つ、持ち帰りでお願いします。",
+    "caption": "落ち着いた大人の話者。店員に注文を伝える短い会話。自然な速さで、最後まで言い切る。",
+    "seed": 4101,
+    "file": "turn-order-full.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "order",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.08,
+    "peak_abs": 0.7851645946502686,
+    "rms": 0.17343387180324984,
+    "generation_seconds": 0.9514303421601653,
+    "sha256": "1cd199edb025dc5e93bb8ef22dc8cb98297ea879ee30e578bef43cd2c0260618",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "backchannel-full",
+    "name": "相手の説明への相づち",
+    "speech_act": "backchannel",
+    "intent": "none",
+    "turn_complete": true,
+    "context": "案内係: 入口を出たら右へ進むと、受付が見えてきます。",
+    "text": "うん、分かりました。",
+    "caption": "親しみのある大人の話者。相手の説明を聞きながら短く相づちを打つ。急がず、自然に言い切る。",
+    "seed": 4102,
+    "file": "turn-backchannel-full.wav",
+    "expected": {
+      "speech_act": "backchannel",
+      "intent": "none",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 4.24,
+    "peak_abs": 0.7609981894493103,
+    "rms": 0.15303109302071377,
+    "generation_seconds": 0.544307358097285,
+    "sha256": "ab9c4c35fa15521b227540689743c15db1fae9a4701cd1c22069fe0789ac844d",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "confirm-full",
+    "name": "内容を確認して承認する",
+    "speech_act": "confirm",
+    "intent": "reservation",
+    "turn_complete": true,
+    "context": "店員: 来週の火曜日、午後六時に二名様でお席をお取りしてよろしいですか。",
+    "text": "はい、その内容でお願いします。",
+    "caption": "穏やかな大人の話者。店員の確認に簡潔に同意する。会話らしい自然な間で、やわらかく言い切る。",
+    "seed": 4103,
+    "file": "turn-confirm-full.wav",
+    "expected": {
+      "speech_act": "confirm",
+      "intent": "reservation",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.16,
+    "peak_abs": 0.6248125433921814,
+    "rms": 0.15614677213260153,
+    "generation_seconds": 0.49432186922058463,
+    "sha256": "5c9569279a31be36d519b3e9e924e187d29e3eba967ed70850915108e4590e51",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "correction-full",
+    "name": "日付を訂正して変更する",
+    "speech_act": "correct",
+    "intent": "change",
+    "turn_complete": true,
+    "context": "店員: ご予約は金曜日の午後三時で承っています。",
+    "text": "あ、すみません。金曜日ではなく、土曜日の午後に変更をお願いします。",
+    "caption": "丁寧な大人の話者。予約内容を思い出して訂正する。最初に短いためらいを置き、最後ははっきり言い切る。",
+    "seed": 4104,
+    "file": "turn-correction-full.wav",
+    "expected": {
+      "speech_act": "correct",
+      "intent": "change",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 6.4,
+    "peak_abs": 0.7035095691680908,
+    "rms": 0.1609569216640416,
+    "generation_seconds": 0.6292441417463124,
+    "sha256": "b73eb1847645cd473f71a8d8296639f6bc38ec17dc628ec843a680d005d44a1c",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "cancel-full",
+    "name": "注文を取り消す",
+    "speech_act": "cancel",
+    "intent": "cancel",
+    "turn_complete": true,
+    "context": "店員: 先ほどのスープとコーヒーのご注文を確認します。",
+    "text": "すみません、スープのほうは取り消してください。",
+    "caption": "落ち着いた大人の話者。注文の一部を丁寧に取り消す。自然な会話の速さで、申し訳なさを少し含めて言い切る。",
+    "seed": 4105,
+    "file": "turn-cancel-full.wav",
+    "expected": {
+      "speech_act": "cancel",
+      "intent": "cancel",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 3.44,
+    "peak_abs": 0.781799852848053,
+    "rms": 0.16470869580552724,
+    "generation_seconds": 0.5071098008193076,
+    "sha256": "24ca56ecc4191620c37834cf5d2f28c51ed7192ee5b6235585fb2dcc0ed38045",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "order-hesitation-unfinished",
+    "name": "迷って途中で止まる注文",
+    "speech_act": "request",
+    "intent": "order",
+    "turn_complete": false,
+    "context": "店員: ご注文が決まりましたら、お声がけください。",
+    "text": "きのこのパスタを一つ、ええと、持ち帰りにするか…",
+    "caption": "迷っている大人の話者。注文を考えながら言いよどみ、答えを決めずに途中で止める。読み上げ調にしない。",
+    "seed": 4106,
+    "file": "turn-order-hesitation-unfinished.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "order",
+      "turn_complete": false
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 5.08,
+    "peak_abs": 0.9923233389854431,
+    "rms": 0.161138963203946,
+    "generation_seconds": 0.5771061358973384,
+    "sha256": "70c4bc0af49280d772c09ea9977f5a1d072f58c71705be9dc36bd1704010ea1e",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "reservation-clause-unfinished",
+    "name": "意味のある途中節で止まる予約",
+    "speech_act": "request",
+    "intent": "reservation",
+    "turn_complete": false,
+    "context": "店員: ご希望の日時と人数をお知らせください。",
+    "text": "あしたの夕方に二人で、もし窓際が…",
+    "caption": "控えめな大人の話者。予約の条件を伝え始めるが、文を完結させずに意味のある途中で止まる。自然な短い間を含める。",
+    "seed": 4107,
+    "file": "turn-reservation-clause-unfinished.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "reservation",
+      "turn_complete": false
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 5.28,
+    "peak_abs": 1.0007697343826294,
+    "rms": 0.13793335127204556,
+    "generation_seconds": 0.5772101040929556,
+    "sha256": "14de328d4e6a4271e671967333fc0c5392d637d0a9876c7b6d0d72a1eb4fa926",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  },
+  {
+    "id": "reservation-full-contrast",
+    "name": "同じ予約条件を最後まで伝える",
+    "speech_act": "request",
+    "intent": "reservation",
+    "turn_complete": true,
+    "context": "店員: ご希望の日時と人数をお知らせください。",
+    "text": "あしたの夕方に二人で、窓際の席を予約したいです。",
+    "caption": "控えめな大人の話者。予約の条件を伝え、最後まで自然に言い切る。文の区切りに短い息継ぎを入れる。",
+    "seed": 4108,
+    "file": "turn-reservation-full-contrast.wav",
+    "expected": {
+      "speech_act": "request",
+      "intent": "reservation",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "reference_embedding": null,
+    "training_excluded": true,
+    "model": "Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_revision": "e863a3a93e652e09afeff3e84823a206a0a60314",
+    "model_card": "https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign",
+    "model_card_license": "MIT",
+    "license_evidence": "spaces/mocovoice-audio-jev/licenses/irodori-model-card.md",
+    "num_steps": 40,
+    "cfg_scale_text": 3.0,
+    "cfg_scale_caption": 2.5,
+    "cfg_guidance_mode": "independent",
+    "watermark_enabled": true,
+    "sample_rate": 48000,
+    "seconds": 5.68,
+    "peak_abs": 0.8254440426826477,
+    "rms": 0.15123665222928692,
+    "generation_seconds": 0.5876368628814816,
+    "sha256": "36a84d9c7061577b438529f4e0882d456f7f3ce773cc9f55248be5fcac1a43a3",
+    "provenance_note": "Original authored text and caption-only VoiceDesign synthesis; no reference audio, embedding, or speaker identity supplied.",
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation. It does not prove a model can make arbitrary floor-taking or turn-taking predictions."
+  }
+]

examples/adult-backchannel-v3.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:424c3152da9a8b1444363ff71574cfe26c855b5fe65e22ab6a553530132bdbe5
+size 334124

examples/adult-cancel-v3.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27a758e3f6db7c2553db937a638f1ddf60e9ad79d3c5f49a8d8097480005dd99
+size 376364

examples/adult-reservation-v3.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e494ae99d12c8fd23ba4741a5a8507942f0e54877b1c43493d5924e0a03c7585
+size 437804

examples/animal-dog-birds-238688.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb229bbe56c08902fb2e1e1d7bad639b49908148e384bdf3a1aa8d9687b19c24
+size 712036

examples/catalog.json

@@ -0,0 +1,169 @@
+[
+  {
+    "name": "予約",
+    "audio": "examples/adult-reservation-v3-telephone.wav",
+    "context": "店員: ご用件をお聞かせください。",
+    "questions": {
+      "intent": {
+        "type": "choice",
+        "instructions": "文脈を踏まえると、この発話の目的は何ですか。",
+        "criteria": {
+          "reservation": "予約を新しく申し込む",
+          "change": "既存の予約や注文の内容を変更する",
+          "cancel": "既存の予約や注文を取り消す",
+          "order": "品物や飲食物を注文する",
+          "none": "予約・注文・変更・取り消しのいずれでもない"
+        }
+      }
+    },
+    "synthetic_audio": true,
+    "synthesis": "Irodori VoiceDesign",
+    "text": "もしもし、あしたの夜七時に、二人で予約できますか。",
+    "expected": {
+      "intent": "reservation"
+    },
+    "telephone_audio": true,
+    "training_excluded": true,
+    "id": "reservation-adult-v3"
+  },
+  {
+    "id": "backchannel-adult-v3",
+    "name": "相づち",
+    "audio": "examples/adult-backchannel-v3.wav",
+    "context": "担当者: こちらの窓口は、土曜日も夕方六時まで開いています。",
+    "text": "ああ、そうなんですね。",
+    "questions": {
+      "speech_act": {
+        "type": "choice",
+        "instructions": "この発話の会話上の働きは何ですか。",
+        "criteria": {
+          "backchannel": "相手の話を聞いていることを示す相づち",
+          "confirm": "確認への同意または承認",
+          "request": "新しい依頼や要求",
+          "correct": "前の内容の訂正や変更",
+          "cancel": "依頼や予定の取り消し",
+          "other": "これらに該当しない発話"
+        }
+      },
+      "intent": {
+        "type": "choice",
+        "instructions": "文脈を踏まえると、この発話の目的は何ですか。",
+        "criteria": {
+          "none": "予約・注文・変更・取り消しのいずれでもない",
+          "reservation": "予約を新しく申し込む",
+          "order": "品物や飲食物を注文する",
+          "change": "既存の予約や注文の内容を変更する",
+          "cancel": "既存の予約や注文を取り消す"
+        }
+      },
+      "turn_complete": {
+        "type": "noul",
+        "instructions": "この発話は意味的に言い終えられていますか。",
+        "criteria": {
+          "false": "発話は意味のある途中で止まっており、話者が言い終えたとは扱わない。",
+          "true": "発話はこの単独の発話として意味的に言い終えられている。"
+        }
+      }
+    },
+    "expected": {
+      "speech_act": "backchannel",
+      "intent": "none",
+      "turn_complete": true
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "training_excluded": true,
+    "annotation_caveat": "turn_complete is an authored semantic-endpoint annotation, not evidence of a true conversational floor-taking decision.",
+    "synthesis": "Irodori VoiceDesign",
+    "telephone_audio": false
+  },
+  {
+    "id": "cancel-adult-v3",
+    "name": "予約キャンセル",
+    "audio": "examples/adult-cancel-v3-telephone.wav",
+    "context": "店員: ご予約についてのご用件をお聞かせください。",
+    "text": "すみません、あしたの予約をキャンセルしたいんですけど。",
+    "questions": {
+      "intent": {
+        "type": "choice",
+        "instructions": "文脈を踏まえると、この発話の目的は何ですか。",
+        "criteria": {
+          "reservation": "予約を新しく申し込む",
+          "change": "既存の予約や注文の内容を変更する",
+          "cancel": "既存の予約や注文を取り消す",
+          "order": "品物や飲食物を注文する",
+          "none": "予約・注文・変更・取り消しのいずれでもない"
+        }
+      }
+    },
+    "expected": {
+      "intent": "cancel"
+    },
+    "synthetic_audio": true,
+    "reference_audio": null,
+    "training_excluded": true,
+    "telephone_audio": true,
+    "synthesis": "Irodori VoiceDesign"
+  },
+  {
+    "id": "music-acoustic-loop-848704",
+    "name": "音楽",
+    "audio": "examples/music-acoustic-loop-848704.wav",
+    "context": "",
+    "questions": {
+      "music_present": {
+        "type": "noul",
+        "instructions": "音声に歌や楽器を含む音楽が含まれますか。",
+        "criteria": {
+          "false": "歌や楽器を含む音楽は含まれない。",
+          "true": "歌や楽器を含む音楽が含まれる。"
+        }
+      }
+    },
+    "expected": {
+      "music_present": true
+    },
+    "synthetic_audio": false,
+    "execution": ["server", "edge"],
+    "source_record_id": "freesound:848704",
+    "training_excluded": true
+  },
+  {
+    "id": "animal-dog-birds-238688",
+    "name": "動物音",
+    "audio": "examples/animal-dog-birds-238688.wav",
+    "context": "",
+    "questions": {
+      "animal_present": {
+        "type": "noul",
+        "instructions": "音声に動物の鳴き声が含まれますか。",
+        "criteria": {
+          "false": "動物音は含まれない。",
+          "true": "動物音が含まれる。"
+        }
+      },
+      "animal_type": {
+        "type": "choice",
+        "instructions": "この録音で聞こえる主な動物の音を選んでください。",
+        "criteria": {
+          "bird": "鳥の音",
+          "cat": "猫の音",
+          "dog": "犬の音",
+          "insect": "昆虫の音",
+          "chicken_and_rooster": "ニワトリ・雄鶏の音",
+          "crow": "カラスの音",
+          "frog": "カエルの音",
+          "cricket": "コオロギの音"
+        }
+      }
+    },
+    "expected": {
+      "animal_present": true,
+      "animal_type": "dog"
+    },
+    "synthetic_audio": false,
+    "execution": ["server", "edge"],
+    "source_record_id": "freesound:238688",
+    "training_excluded": true
+  }
+]

examples/irodori-cancel.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25945f9ea6923cb47a1b008e0dbe935e4ee30bb3a30a1a030c12817b6a693b01
+size 468524

examples/irodori-change.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c3fe27ad38a581e1f545c6cc8b18f8dd7a75b24e7c2737952e3b0e88ae589c2
+size 453164

examples/irodori-reservation.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f8ee91977655e119b26d3d7d35e9b414e62f1ef9c6b2c46097f010442f61fd9
+size 384044

examples/music-639.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2637513ffafbae9b74da216e10b4b4dd99d966165db89284ce9da38593727cfa
+size 172562

examples/music-acoustic-loop-848704.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c851b6cad67ac9ef08f5c634efb90f001670f5ed454320c276310bfe9b7c020a
+size 513518

examples/natural-cancel-b.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52f0ed8dfd7ef2cbdd0ce67533f4f1acf2fe7ced93c9e34aa5d49957609f35dc
+size 860204

examples/natural-change-a.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1a9828f1beada7c90182df61d30872a5edc056bef833219fd53fc48ee243a2d
+size 637484

examples/natural-reservation-a-telephone.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c30790ee4ef54eb4e12f7ea58ff1cd2d65d399560f174e61d5a2451b0aa7f77
+size 135084

examples/natural-reservation-a.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf9ce6d83ae7fe8961f97f4cd95a7d6be1ca29d2c373cc77c96b7a41d7d719d2
+size 810284

examples/reservation.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1b4a56e9a2068f092515b564a06bde60a329b60389fe09f5101a6f77aab9fde
+size 137324

examples/turn-backchannel-full.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab9c4c35fa15521b227540689743c15db1fae9a4701cd1c22069fe0789ac844d
+size 407084

examples/turn-cancel-full.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24ca56ecc4191620c37834cf5d2f28c51ed7192ee5b6235585fb2dcc0ed38045
+size 330284

examples/turn-confirm-full.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c9569279a31be36d519b3e9e924e187d29e3eba967ed70850915108e4590e51
+size 303404

examples/turn-correction-full.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b73eb1847645cd473f71a8d8296639f6bc38ec17dc628ec843a680d005d44a1c
+size 614444

examples/turn-order-full.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cd199edb025dc5e93bb8ef22dc8cb98297ea879ee30e578bef43cd2c0260618
+size 391724

examples/turn-order-hesitation-unfinished.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70c4bc0af49280d772c09ea9977f5a1d072f58c71705be9dc36bd1704010ea1e
+size 487724

examples/turn-reservation-clause-unfinished.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14de328d4e6a4271e671967333fc0c5392d637d0a9876c7b6d0d72a1eb4fa926
+size 506924

examples/turn-reservation-full-contrast.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36a84d9c7061577b438529f4e0882d456f7f3ce773cc9f55248be5fcac1a43a3
+size 545324

fonts/LINESeedJP_OTF_Bd.woff2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:512c1b416bb24e281e182c5a70f3909a8ebd0b5494082f9b7a7c157452c7026f
+size 2156508

fonts/LINESeedJP_OTF_Rg.woff2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75a41f51f4b2d3172fde7a45daf544a58df58f5791d996c3cb3fbf855f67c7a8
+size 2104408

index.html

@@ -0,0 +1,37 @@
+<!doctype html>
+<html lang="ja">
+<head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>Audio Decision Model by mocomoco inc.</title><link rel="icon" href="./favicon.svg" type="image/svg+xml"><link rel="preload" href="./mocomoco-inc-logo.svg" as="image" type="image/svg+xml"><link rel="preload" href="./fonts/LINESeedJP_OTF_Rg.woff2" as="font" type="font/woff2" crossorigin><link rel="preload" href="./fonts/LINESeedJP_OTF_Bd.woff2" as="font" type="font/woff2" crossorigin><link rel="stylesheet" href="./styles.css?v=20260921-title-width-v1"></head>
+<body>
+<div id="startup-screen" class="startup-screen" role="status" aria-label="Audio Decision Modelを読み込んでいます"><div class="startup-stage"><span class="startup-spinner" aria-hidden="true"></span><div class="startup-brand"><strong>Audio Decision Model</strong><span>by <img src="./mocomoco-inc-logo.svg" alt="mocomoco inc." width="217" height="33"></span></div></div></div>
+<div id="mode-color-edge" aria-hidden="true"></div>
+<header class="topbar"><div class="product-brand"><h1 class="product-title"><a href="./"><span id="product-name">Audio Decision Model</span> <span id="edge-mode-label" class="edge-mode-label" hidden>Edge AI</span> <span class="product-byline">by <img src="./mocomoco-inc-logo.svg" alt="mocomoco inc." width="217" height="33"></span></a></h1><p class="product-inspiration">Inspired by <a href="https://typesafe.ai/blog/introducing-system-one-models-and-jev" target="_blank" rel="noopener noreferrer">TypeSafe AI’s Jev</a></p><p class="product-description">音声入力に対する判断を高速に返すAIです。言語情報・非言語情報の判断形式を指定できます。</p></div><a class="hf-link hf-link-top" href="https://huggingface.co/mocomoco-inc/mocomoco-AudioJev-AudioDecisionModel" target="_blank" rel="noopener noreferrer"><span aria-hidden="true">🤗</span>Available on Hugging Face<svg aria-hidden="true" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6"><path d="M8 5h11v11M19 5 5 19"/></svg></a></header>
+<main class="page">
+<div class="toolbar"><div class="execution-picker"><span class="execution-label" aria-hidden="true">AIの処理場所</span><fieldset id="execution-toggle" class="execution-toggle"><legend class="sr-only">AIの処理場所</legend><label class="execution-option"><input type="radio" name="execution" value="server" checked><span><svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.7" aria-hidden="true"><path d="M7 19a5 5 0 0 1-.7-9.95A6 6 0 0 1 18 8a5.5 5.5 0 0 1 .5 11H7Z"/></svg>クラウド</span></label><label class="execution-option"><input type="radio" name="execution" value="edge"><span><svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.7" aria-hidden="true"><rect x="3" y="4" width="18" height="13" rx="2"/><path d="M8 21h8m-4-4v4M3 8h18"/></svg>Edge AI</span></label></fieldset></div><select id="execution-mode" hidden aria-hidden="true"><option value="server">クラウド · CPU</option><option value="edge">Edge AI · CPU</option></select><div id="mode-announcement" class="mode-description" role="status" aria-live="polite"><p id="mode-announcement-text"></p></div><span id="backend-wrap" hidden>WASM / INT8</span><span id="execution-kicker" hidden></span><span id="model-title" hidden></span></div>
+<div class="workspace">
+<div class="inputs">
+<section class="section audio-section"><div class="section-heading"><h2>音声</h2><span class="hint">最大 30 秒</span></div>
+<div class="recorder" id="recorder" aria-label="録音、ファイル選択、または音声ファイルのドラッグ＆ドロップ">
+<div class="recorder-meta"><span id="record-state"><span class="ready-dot"></span>マイクで録音</span><span id="record-timer">0:00 <span>/ 0:30</span></span></div>
+<div class="waveform"><canvas id="record-waveform" aria-hidden="true"></canvas></div>
+<div class="audio-controls"><button id="record" class="record-button"><svg width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" aria-hidden="true"><rect x="9" y="2" width="6" height="12" rx="3"/><path d="M5 10v2a7 7 0 0 0 14 0v-2M12 19v3M8 22h8"/></svg>録音をはじめる</button><span class="upload-divider">または</span><label class="file-button" for="audio-file">ファイルを選ぶ<input id="audio-file" type="file" accept="audio/*"></label></div>
+</div><div id="examples-section" hidden><span class="hint">サンプル</span><div id="examples" class="examples"></div></div><audio id="audio-preview" hidden preload="metadata"></audio><div id="audio-player" class="audio-player" hidden><div class="player-heading"><span>音声プレビュー</span><span id="player-time">0:00 / 0:00</span></div><div class="player-controls"><button id="player-play" class="player-play" type="button" aria-label="再生"></button><div class="player-track"><canvas id="player-waveform" aria-hidden="true"></canvas><input id="player-seek" type="range" min="0" max="1" step="0.01" value="0" aria-label="再生位置"></div><select id="player-speed" aria-label="再生速度"><option value="1">1×</option><option value="1.25">1.25×</option><option value="1.5">1.5×</option><option value="2">2×</option></select></div></div><p id="sample-transcript" class="sample-transcript" hidden></p><p id="audio-status" class="hint"></p>
+<details class="context-details"><summary>文脈・補足 <span class="hint">任意</span></summary><textarea id="context" maxlength="2000" rows="3" placeholder="直前の会話など"></textarea><span class="hint"><span id="context-count">0</span> / 2000</span></details>
+</section>
+<section class="section questions-section"><div class="section-heading"><h2>質問</h2><select id="preset" aria-label="プリセット"><option value="identity">予約・注文の意図</option><option value="language">話されている言語</option><option value="service">音のカテゴリ</option><option value="classification">音カテゴリ数</option><option value="turntaking">発話の役割</option><option value="animals">動物の種類</option></select></div>
+<label class="question-selector">編集する質問<select id="existing-question"></select></label>
+<form id="question-builder" class="builder">
+<label>質問文<textarea id="question-instructions" rows="3" required></textarea></label>
+<label>回答形式<select id="question-type"><option value="noul">Yes / No</option><option value="choice">選択</option><option value="score">数値</option></select></label>
+<div id="criteria-panel" class="criteria-fields" hidden>
+<div class="criteria-fields-heading"><span id="criteria-title">選択肢</span><span id="criteria-limit" class="hint"></span></div>
+<div id="criteria-items" class="criteria-items"></div>
+<button id="add-criterion" class="add-criterion" type="button">＋ 選択肢を追加</button>
+</div>
+</form>
+<details class="question-advanced"><summary>詳細設定</summary><div class="advanced-fields"><label>質問ID <span class="hint">結果JSONのキー</span><input id="question-name" required></label><label>質問セット JSON<textarea id="questions-json" rows="16" spellcheck="false"></textarea></label></div></details><p id="question-error" class="error" role="alert"></p>
+</section>
+
+</div>
+<aside class="output"><div class="section-heading"><h2>結果</h2></div><div class="run-bar"><button id="analyze" class="primary">実行する</button><span id="analysis-status" role="status">音声を選んでください。</span></div><div id="summary" class="summary is-empty">音声を選んで解析すると、ここに結果が表示されます。</div><details class="result-details" open><summary>結果 JSON</summary><pre id="result-json">{}</pre></details>
+</aside>
+</div></main><footer class="site-footer"><div class="operator-block"><div class="operator"><span>企業情報</span><a class="operator-logo" href="https://mocomoco.ai" target="_blank" rel="noopener noreferrer"><img src="./mocomoco-inc-logo.svg" alt="mocomoco inc." width="217" height="33" loading="lazy"></a></div><p class="references">参考にしているAI：<a href="https://docs.typesafe.ai/introduction" target="_blank" rel="noopener noreferrer">Jev / TypeSafe AI</a> · <a href="https://github.com/TheoLeeCJ/SemIf" target="_blank" rel="noopener noreferrer">OpenJev / SemIf</a><br>Audio Decision Modelは、これらを参考にmocomoco inc.が研究開発する音声向けの独立した実装です。</p></div><a class="hf-link" href="https://huggingface.co/mocomoco-inc/mocomoco-AudioJev-AudioDecisionModel" target="_blank" rel="noopener noreferrer"><span aria-hidden="true">🤗</span>Available on Hugging Face<svg aria-hidden="true" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6"><path d="M8 5h11v11M19 5 5 19"/></svg></a></footer><dialog id="edge-intro" class="edge-intro" aria-labelledby="edge-intro-title" aria-describedby="edge-intro-description"><button id="edge-intro-close" class="dialog-close" type="button" aria-label="説明を閉じる" autofocus><svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" aria-hidden="true"><path d="m6 6 12 12M18 6 6 18"/></svg></button><svg class="edge-intro-illustration" viewBox="0 0 280 110" fill="none" role="img" aria-label="音声を端末の中で処理し、外へ送信しないイメージ"><rect x="65" y="12" width="150" height="78" rx="9" fill="#162d40" stroke="#7293aa" stroke-width="2"/><rect x="74" y="21" width="132" height="60" rx="4" fill="#101b2a"/><path d="M53 91h174l-8 9H61l-8-9Z" fill="#203c51" stroke="#7293aa" stroke-width="2" stroke-linejoin="round"/><path d="M128 94h24" stroke="#7293aa" stroke-width="2" stroke-linecap="round"/><g stroke="#69d2ff" stroke-width="3" stroke-linecap="round"><path d="M91 47v10m7-18v26m7-31v36m7-27v18m7-13v8"/></g><path d="M132 52h20m-5-5 5 5-5 5" stroke="#7293aa" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"/><rect x="163" y="45" width="27" height="22" rx="5" fill="#25485e" stroke="#69d2ff" stroke-width="2"/><path d="M170 45v-6a6.5 6.5 0 0 1 13 0v6" stroke="#69d2ff" stroke-width="2"/><path d="m171 56 4 4 7-8" stroke="#a0e4ff" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"/></svg><h2 id="edge-intro-title">Edge AI</h2><div id="edge-intro-description"><p>データをクラウドに送らず、この端末内でJevを実行します。外部送信不要で、オフラインでも利用できます。初回のみ「モデルを読み込む」からモデルをダウンロードします。</p></div></dialog><script type="module" src="./ui.js?v=20260921-preload-all-v2"></script></body></html>

inference-worker.js

@@ -0,0 +1,435 @@
+import * as ort from './vendor/ort.webgpu.bundle.min.mjs';
+import {Tokenizer} from './vendor/tokenizers.mjs';
+import {WhisperFeatureExtractor, ClapFeatureExtractor} from './vendor/transformers.web.min.js';
+
+ort.env.wasm.wasmPaths = new URL('./vendor/', import.meta.url).href;
+ort.env.wasm.numThreads = 1;
+ort.env.wasm.proxy = false;
+
+let model = null;
+let loading = null;
+let languageLoading = null;
+let soundLoading = null;
+let currentId = null;
+const textCache = new Map();
+const truncatedTexts = new Set();
+const progress = message => postMessage({id: currentId, progress: message});
+
+const BROAD = ['animal', 'nature', 'music', 'environment'];
+const ANIMAL8 = ['dog', 'cat', 'bird', 'insect', 'chicken_and_rooster', 'crow', 'frog', 'cricket'];
+const ALIASES = {
+  animal: ['animal sound', 'animal', '動物音', '動物の音', '動物の鳴き声'],
+  nature: ['nature sound', 'nature', '自然音', '自然の音'],
+  music: ['music', 'musical', '音楽'],
+  environment: ['environmental sound', 'environment sound', 'environment', '環境音', '環境の音'],
+  dog: ['dog', '犬', 'いぬ', '犬の鳴き声'], cat: ['cat', '猫', 'ねこ', '猫の鳴き声'],
+  bird: ['bird', '鳥', 'とり', '鳥の鳴き声'], insect: ['insect', '虫', '昆虫', '虫の音'],
+  chicken_and_rooster: ['chicken', 'rooster', 'hen', '鶏', 'にわとり', 'ニワトリ', '雄鶏'],
+  crow: ['crow', 'カラス', 'からす'], frog: ['frog', 'カエル', 'かえる'],
+  cricket: ['cricket', 'コオロギ', 'こおろぎ'],
+};
+const OR_CUES = ['または', 'もしくは', 'あるいは', 'いずれか', 'or', 'either'];
+const NEGATIVE_CUES = ['含まれない', '聞こえない', '存在しない', '該当しない', '満たさない', 'ではない', 'なし', 'ない', 'without', 'absent', 'not', 'no '];
+
+async function asset(path) {
+  const response = await fetch(new URL(path, import.meta.url));
+  if (!response.ok) throw new Error(`モデルファイルの取得に失敗しました (${response.status})。Spaceへのアクセス権を確認してください。`);
+  return response;
+}
+
+async function modelBytes(path, label) {
+  progress(`${label}を端末に読み込み中…`);
+  const response = await asset(path);
+  const total = Number(response.headers.get('Content-Length') || 0);
+  if (!response.body) return new Uint8Array(await response.arrayBuffer());
+  const reader = response.body.getReader();
+  const chunks = []; let received = 0; let last = performance.now();
+  while (true) {
+    const {done, value} = await reader.read(); if (done) break;
+    chunks.push(value); received += value.length;
+    if (performance.now() - last > 400) {
+      progress(`${label}: ${(received / 1e6).toFixed(0)} MB${total ? ' / ' + (total / 1e6).toFixed(0) + ' MB' : ''}`);
+      last = performance.now();
+    }
+  }
+  const result = new Uint8Array(received); let offset = 0;
+  for (const chunk of chunks) {result.set(chunk, offset); offset += chunk.length;}
+  return result;
+}
+
+async function createSession(path, label) {
+  return ort.InferenceSession.create(await modelBytes(path, label), {
+    executionProviders: ['wasm'], graphOptimizationLevel: 'all',
+  });
+}
+
+async function load() {
+  if (model?.audio && model?.head && model?.sound) return status();
+  if (loading) return loading;
+  loading = (async () => {
+    if (!model) {
+      const manifest = await (await asset('./models/manifest.json')).json();
+      progress('Edge AIモデルを端末に読み込みます。');
+      const text = await createSession('./models/text_q8.onnx', '質問モデル');
+      const routerCore = await createSession('./models/modality_router_core_fp32.onnx?v=20260921-router-ensemble-web-v1', 'ルーティングモデル 1/2');
+      const routerNatural = await createSession('./models/modality_router_natural_fp32.onnx?v=20260921-router-ensemble-web-v1', 'ルーティングモデル 2/2');
+      const [tokenizerJson, tokenizerConfig, qwenConfig, clapConfig] = await Promise.all([
+        'tokenizer.json', 'tokenizer_config.json',
+      ].map(async name => (await asset('./models/tokenizer/' + name)).json()).concat([
+        asset('./models/qwen_preprocessor_config.json').then(response => response.json()),
+        asset('./models/clap_preprocessor_config.json').then(response => response.json()),
+      ]));
+      model = {
+        audio: null, text, head: null, routers: [routerCore, routerNatural], sound: null,
+        tokenizer: new Tokenizer(tokenizerJson, tokenizerConfig),
+        qwenExtractor: new WhisperFeatureExtractor(qwenConfig),
+        clapExtractor: new ClapFeatureExtractor(clapConfig),
+        manifest, backend: 'wasm', precision: 'int8',
+      };
+    }
+    await loadLanguage();
+    await loadSound();
+    progress('モデルの準備ができました。データはこの端末内だけで処理されます。');
+    return status();
+  })();
+  try {return await loading;} finally {loading = null;}
+}
+
+async function loadLanguage() {
+  if (model.audio && model.head) return {audio: model.audio, head: model.head};
+  if (languageLoading) return languageLoading;
+  languageLoading = (async () => {
+    progress('発話モデルを端末に読み込み中…');
+    const audio = await createSession('./models/qwen_audio_q8.onnx', '発話モデル');
+    const head = await createSession('./models/qwen_typed_head_q8.onnx?v=20260921-noulmatch-v4', '発話の判断モデル');
+    model.audio = audio; model.head = head;
+    return {audio, head};
+  })();
+  try {return await languageLoading;} finally {languageLoading = null;}
+}
+
+async function loadSound() {
+  if (model.sound) return model.sound;
+  if (soundLoading) return soundLoading;
+  soundLoading = (async () => {
+    progress('非言語モデルを端末に読み込み中…');
+    const audio = await createSession('./models/clap_audio_q8.onnx', '非言語モデル');
+    const heads = await createSession('./models/clap_heads_q8.onnx', '非言語の判断モデル');
+    model.sound = {audio, heads};
+    return model.sound;
+  })();
+  try {return await soundLoading;} finally {soundLoading = null;}
+}
+
+function status() {
+  return {ready: true, backend: model.backend, precision: model.precision,
+    model: model.manifest.model, on_device: true, language_ready: Boolean(model.audio), sound_ready: Boolean(model.sound)};
+}
+
+function optionsFor(q) {
+  if (!q || typeof q !== 'object' || typeof q.instructions !== 'string' || !q.instructions.trim()) throw new Error('各質問にinstructions（質問文）を指定してください。');
+  if (q.instructions.length > 2000) throw new Error('質問文は2000文字以内にしてください。');
+  let pairs;
+  if (q.type === 'noul') {
+    const c = q.criteria ?? {false: 'いいえ。その条件を満たさない。', true: 'はい。その条件を満たす。'};
+    if (!c || Array.isArray(c) || Object.keys(c).sort().join(',') !== 'false,true') throw new Error('Noulのcriteriaにはfalseとtrueを指定してください。');
+    pairs = [['false', c.false], ['true', c.true]];
+  } else if (q.type === 'choice') {
+    if (!q.criteria || typeof q.criteria !== 'object' || Array.isArray(q.criteria)) throw new Error('Choiceのcriteriaは候補IDと説明のオブジェクトです。');
+    pairs = Object.entries(q.criteria);
+    if (pairs.length < 2 || pairs.length > 32) throw new Error('Choiceの候補数は2〜32個です。');
+  } else if (q.type === 'score') {
+    if (!Array.isArray(q.criteria) || q.criteria.length < 2 || q.criteria.length > 10) throw new Error('Scoreのcriteriaは2〜10段階の説明を順に並べた配列です。');
+    pairs = q.criteria.map((value, index) => [String(index), value]);
+  } else throw new Error('typeはnoul・choice・scoreのいずれかです。');
+  if (pairs.some(([key, value]) => !key || typeof value !== 'string' || !value.trim() || value.length > 1000)) throw new Error('各候補には空でない説明を1000文字以内で指定してください。');
+  return pairs;
+}
+
+function contains(text, alias) {
+  const value = text.toLocaleLowerCase().replace(/\s+/g, ' ').trim();
+  const target = alias.toLocaleLowerCase().replace(/\s+/g, ' ').trim();
+  if (/^[\x00-\x7F]+$/.test(target) && /[a-z]/.test(target)) {
+    const escaped = target.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+    return new RegExp(`(^|[^a-z])${escaped}([^a-z]|$)`).test(value);
+  }
+  return value.includes(target);
+}
+
+function labelsIn(text, allowed) {
+  return allowed.filter(label => ALIASES[label].some(alias => contains(text, alias)));
+}
+
+function hasAny(text, cues) { return cues.some(cue => contains(text, cue)); }
+function operatorFor(text) { return hasAny(text, OR_CUES) ? 'or' : 'and'; }
+function isNegative(text) { return hasAny(text, NEGATIVE_CUES); }
+
+function softmax(values) {
+  const maximum = Math.max(...values);
+  const exponentials = values.map(value => Math.exp(value - maximum));
+  const total = exponentials.reduce((sum, value) => sum + value, 0);
+  return exponentials.map(value => value / total);
+}
+
+async function learnedRoute(contextVector, questionVector, optionVectors, count) {
+  const inputs = {
+    context: new ort.Tensor('float32', contextVector, [1, 256]),
+    question: new ort.Tensor('float32', questionVector, [1, 256]),
+    options: new ort.Tensor('float32', optionVectors, [1, count, 256]),
+  };
+  const core = await model.routers[0].run(inputs);
+  const natural = await model.routers[1].run(inputs);
+  const mix = (left, right) => Array.from(left, (value, index) => 0.4 * value + 0.6 * right[index]);
+  const questionLogits = mix(core.question_logits.data, natural.question_logits.data);
+  const optionLogits = mix(core.option_logits.data, natural.option_logits.data);
+  const weights = softmax(questionLogits);
+  const optionWeights = Array.from({length: count}, (_, index) =>
+    softmax(optionLogits.slice(index * 3, index * 3 + 3)));
+  const entropy = -weights.reduce((sum, value) => sum + value * Math.log(Math.max(value, 1e-12)), 0) / Math.log(3);
+  const both = weights[2] >= 0.10 || entropy >= 0.40;
+  const mode = both ? 'joint' : (weights[0] >= weights[1] ? 'language' : 'sound');
+  return {mode, weights, optionWeights, entropy};
+}
+
+function publicRoute(route) {
+  const experts = route.mode === 'language' ? ['speech'] : route.mode === 'sound' ? ['nonverbal'] : ['speech', 'nonverbal'];
+  const value = {mode: route.mode, experts, method: 'learned_sparse_router_ensemble_v1', entropy: Number(route.entropy.toFixed(4)),
+    weights: Object.fromEntries(['speech', 'nonverbal', 'joint'].map((name, index) => [name, Number(route.weights[index].toFixed(4))]))};
+  if (route.mode === 'joint') value.fusion = 'learned_option_weighted_logit_fusion_v1';
+  return value;
+}
+
+function typedAnswer(question, pairs, probabilities) {
+  const total = probabilities.reduce((sum, value) => sum + value, 0);
+  const ps = probabilities.map(value => value / total);
+  const entropy = -ps.reduce((sum, value) => sum + (value ? value * Math.log(value) : 0), 0) / Math.log(ps.length);
+  const answer = {type: question.type, probabilities: Object.fromEntries(pairs.map(([key], index) => [key, ps[index]])),
+    confidence: Math.max(0, Math.min(1, 1 - entropy)), confidence_kind: 'normalized_entropy_not_calibrated'};
+  if (question.type === 'noul') answer.noul = ps[1];
+  else if (question.type === 'choice') answer.choice = pairs[ps.indexOf(Math.max(...ps))][0];
+  else {answer.score = ps.reduce((sum, value, index) => sum + index * value, 0); answer.legend = Object.fromEntries(pairs);}
+  return answer;
+}
+
+async function textVector(value, warnings) {
+  if (textCache.has(value)) {
+    if (truncatedTexts.has(value)) warnings.add('長い文はモデルの上限128トークンに切り詰めました。');
+    return textCache.get(value);
+  }
+  const tokenized = model.tokenizer.encode(value, {add_special_tokens: true});
+  let tokens = Array.from(tokenized.ids);
+  if (tokens.length > 128) {truncatedTexts.add(value); tokens = [...tokens.slice(0, 127), 2]; warnings.add('長い文はモデルの上限128トークンに切り詰めました。');}
+  const ids = new BigInt64Array(128); ids.fill(3n);
+  const mask = new BigInt64Array(128);
+  tokens.forEach((token, index) => {ids[index] = BigInt(token); mask[index] = 1n;});
+  const output = await model.text.run({input_ids: new ort.Tensor('int64', ids, [1, 128]), attention_mask: new ort.Tensor('int64', mask, [1, 128])});
+  const vector = Float32Array.from(output.features.data); textCache.set(value, vector);
+  if (textCache.size > 512) {const key = textCache.keys().next().value; textCache.delete(key); truncatedTexts.delete(key);}
+  return vector;
+}
+
+function qwenWindows(samples) {
+  const width = 80000, hop = 40000;
+  const last = Math.max(0, samples.length - width); const starts = [];
+  for (let index = 0; index <= last; index += hop) starts.push(index);
+  if (!starts.length || starts.at(-1) !== last) starts.push(last);
+  return starts.map(start => samples.slice(start, Math.min(start + width, samples.length)));
+}
+
+async function qwenFeatures(samples) {
+  const language = await loadLanguage();
+  const windows = qwenWindows(samples); const features = new Float32Array(windows.length * 2048);
+  for (let index = 0; index < windows.length; index += 1) {
+    progress(`端末内で発話を分析中 ${index + 1}/${windows.length}`);
+    const extracted = await model.qwenExtractor(windows[index], {max_length: 80000});
+    const mel = Float32Array.from(extracted.input_features.data);
+    const output = await language.audio.run({input_features: new ort.Tensor('float32', mel, [128, 500])});
+    features.set(output.pooled_features.data, index * 2048);
+  }
+  return {features, count: windows.length};
+}
+
+function clapStarts(length) {
+  const count = Math.min(3, Math.max(1, Math.ceil(length / 480000)));
+  const last = Math.max(0, length - 480000);
+  if (count === 1) return [0];
+  return Array.from({length: count}, (_, index) => Math.trunc(index * last / (count - 1)));
+}
+
+async function clapScores(samples48k) {
+  const sound = await loadSound();
+  const starts = clapStarts(samples48k.length); const embeddings = [];
+  for (let index = 0; index < starts.length; index += 1) {
+    progress(`端末内で音を分析中 ${index + 1}/${starts.length}`);
+    const wave = samples48k.slice(starts[index], Math.min(starts[index] + 480000, samples48k.length));
+    const extracted = await model.clapExtractor(wave);
+    const mel = Float32Array.from(extracted.input_features.data);
+    const output = await sound.audio.run({input_features: new ort.Tensor('float32', mel, [1, 1, 1001, 64])});
+    embeddings.push(Float32Array.from(output.audio_embeds.data));
+  }
+  const feature = new Float32Array(1024);
+  for (let dimension = 0; dimension < 512; dimension += 1) {
+    let sum = 0, maximum = -Infinity;
+    for (const embedding of embeddings) {sum += embedding[dimension]; maximum = Math.max(maximum, embedding[dimension]);}
+    feature[dimension] = sum / embeddings.length; feature[512 + dimension] = maximum;
+  }
+  const output = await sound.heads.run({features: new ort.Tensor('float32', feature, [1, 1024])});
+  const sigmoid = values => Array.from(values, value => 1 / (1 + Math.exp(-value)));
+  const broadValues = sigmoid(output.broad_logits.data);
+  const animalValues = sigmoid(output.animal_logits.data);
+  return {broad: Object.fromEntries(['animal', 'nature', 'music', 'environment', 'dog', 'cat', 'bird', 'insect']
+      .map((label, index) => [label, broadValues[index]])),
+    animal8: Object.fromEntries(ANIMAL8.map((label, index) => [label, animalValues[index]])),
+    windows: starts.length};
+}
+
+function allSoundLabels(text) { return [...labelsIn(text, BROAD), ...labelsIn(text, ANIMAL8)]; }
+function labelScore(label, scores) {
+  const value = Number(scores[ANIMAL8.includes(label) ? 'animal8' : 'broad'][label]);
+  if (!Number.isFinite(value)) throw new Error('音の判定値が不正です。');
+  return Math.max(0, Math.min(1, value));
+}
+function combineProbabilities(values, operator) {
+  if (!values.length) throw new Error('対応する音の種類を質問に指定してください。');
+  return operator === 'or' ? 1 - values.reduce((product, value) => product * (1 - value), 1)
+    : values.reduce((product, value) => product * value, 1);
+}
+function soundProbability(text, labels, scores, operator) {
+  const value = combineProbabilities(labels.map(label => labelScore(label, scores)), operator);
+  return isNegative(text) ? 1 - value : value;
+}
+function poissonBinomial(values) {
+  let distribution = [1];
+  for (const value of values) {
+    const next = new Array(distribution.length + 1).fill(0);
+    distribution.forEach((probability, count) => {
+      next[count] += probability * (1 - value); next[count + 1] += probability * value;
+    });
+    distribution = next;
+  }
+  return distribution;
+}
+function soundLikelihoods(question, pairs, scores) {
+  const instructionLabels = allSoundLabels(question.instructions);
+  const optionLabels = pairs.map(([, description]) => allSoundLabels(description));
+  if (question.type === 'score' && instructionLabels.length && pairs.length === instructionLabels.length + 1) {
+    return poissonBinomial(instructionLabels.map(label => labelScore(label, scores)));
+  }
+  const values = pairs.map(([, description], index) => optionLabels[index].length
+    ? soundProbability(description, optionLabels[index], scores, operatorFor(description)) : null);
+  if (values.some(value => value !== null)) return values.map(value => value === null ? 1 : value);
+  const labels = instructionLabels.length ? instructionLabels
+    : allSoundLabels([question.instructions, ...pairs.map(([, value]) => value)].join('\n'));
+  if (!labels.length) throw new Error('質問に対応する音の種類を指定してください。');
+  const probability = soundProbability(question.instructions, labels, scores, operatorFor(question.instructions));
+  if (pairs.length === 2) return [1 - probability, probability];
+  if (question.type === 'score') {
+    const maximum = pairs.length - 1;
+    return pairs.map((_, index) => (1 - probability) * (1 - index / maximum) + probability * (index / maximum));
+  }
+  throw new Error('音のChoiceでは、各選択肢が表す音を説明してください。');
+}
+function soundAnswer(question, pairs, scores) {
+  if (question.type === 'noul') {
+    const labels = allSoundLabels(question.instructions);
+    const fallback = labels.length ? labels : allSoundLabels([question.instructions, ...pairs.map(([, value]) => value)].join('\n'));
+    const value = soundProbability(question.instructions, fallback, scores, operatorFor(question.instructions));
+    return typedAnswer(question, pairs, [1 - value, value]);
+  }
+  return typedAnswer(question, pairs, soundLikelihoods(question, pairs, scores));
+}
+function fuseAnswer(question, pairs, languageAnswer, scores, route) {
+  const language = pairs.map(([key]) => Number(languageAnswer.probabilities[key]));
+  const acoustic = soundLikelihoods(question, pairs, scores);
+  const combined = pairs.map((_, index) => {
+    const [speechWeight, nonverbalWeight, jointWeight] = route.optionWeights[index];
+    const languageWeight = speechWeight + 0.5 * jointWeight;
+    const soundWeight = nonverbalWeight + 0.5 * jointWeight;
+    return languageWeight * Math.log(Math.max(language[index], 1e-8))
+      + soundWeight * Math.log(Math.max(acoustic[index], 1e-8));
+  });
+  const answer = typedAnswer(question, pairs, softmax(combined));
+  answer.fusion = {method: 'learned_option_weighted_logit_fusion_v1'};
+  return answer;
+}
+
+async function analyze(samples, samples48k, context, questions) {
+  if (!(samples instanceof Float32Array)) samples = new Float32Array(samples);
+  if (!(samples48k instanceof Float32Array)) samples48k = new Float32Array(samples48k || []);
+  if (samples.length < 4000 || samples.length > 480000 || samples.some(value => !Number.isFinite(value))) throw new Error('音声は0.25秒〜30秒の範囲で入力してください。');
+  if (samples48k.length < 12000 || samples48k.length > 1440000 || samples48k.some(value => !Number.isFinite(value))) throw new Error('音声の端��内変換に失敗しました。音声を選び直してください。');
+  if (typeof context !== 'string' || context.length > 2000) throw new Error('文脈は2000文字以内にしてください。');
+  if (!questions || typeof questions !== 'object' || Array.isArray(questions)) throw new Error('質問セットにはJSONオブジェクトを指定してください。');
+  const entries = Object.entries(questions);
+  if (!entries.length || entries.length > 16) throw new Error('質問数は1〜16個にしてください。');
+  const pairs = entries.map(([, question]) => optionsFor(question));
+  await load();
+  const started = performance.now(); const warnings = new Set(); const answers = Object.create(null);
+  const contextVector = await textVector(context, warnings);
+  const questionVectors = []; const optionVectorsByQuestion = []; const routes = [];
+  for (let index = 0; index < entries.length; index += 1) {
+    const [, question] = entries[index]; const choices = pairs[index];
+    const questionVector = await textVector(question.instructions, warnings);
+    const optionVectors = new Float32Array(choices.length * 256);
+    for (let option = 0; option < choices.length; option += 1) {
+      optionVectors.set(await textVector(choices[option][1], warnings), option * 256);
+    }
+    questionVectors.push(questionVector); optionVectorsByQuestion.push(optionVectors);
+    routes.push(await learnedRoute(contextVector, questionVector, optionVectors, choices.length));
+  }
+
+  let qwenCount = 0, qwenMs = 0;
+  if (routes.some(route => route.mode === 'language' || route.mode === 'joint')) {
+    const qwenStarted = performance.now();
+    const audio = await qwenFeatures(samples); qwenCount = audio.count;
+    for (let index = 0; index < entries.length; index += 1) {
+      if (routes[index].mode === 'sound') continue;
+      const [name, question] = entries[index]; const choices = pairs[index];
+      progress(`質問を判定中 ${index + 1}/${entries.length}`);
+      const output = await model.head.run({
+        audio: new ort.Tensor('float32', audio.features, [1, audio.count, 2048]),
+        audio_mask: new ort.Tensor('bool', new Uint8Array(audio.count).fill(1), [1, audio.count]),
+        context: new ort.Tensor('float32', contextVector, [1, 256]),
+        question: new ort.Tensor('float32', questionVectors[index], [1, 256]),
+        options: new ort.Tensor('float32', optionVectorsByQuestion[index], [1, choices.length, 256]),
+        option_mask: new ort.Tensor('bool', new Uint8Array(choices.length).fill(1), [1, choices.length]),
+        is_noul: new ort.Tensor('bool', new Uint8Array([question.type === 'noul' ? 1 : 0]), [1]),
+      });
+      const logits = Array.from(output.logits.data); const maximum = Math.max(...logits);
+      const exponentials = logits.map(value => Math.exp(value - maximum)); const total = exponentials.reduce((sum, value) => sum + value, 0);
+      answers[name] = typedAnswer(question, choices, exponentials.map(value => value / total));
+    }
+    qwenMs = performance.now() - qwenStarted;
+  }
+
+  let clapWindows = 0, clapMs = 0;
+  if (routes.some(route => route.mode === 'sound' || route.mode === 'joint')) {
+    const clapStarted = performance.now(); const scores = await clapScores(samples48k); clapWindows = scores.windows;
+    for (let index = 0; index < entries.length; index += 1) {
+      const route = routes[index]; if (route.mode === 'language') continue;
+      const [name, question] = entries[index];
+      answers[name] = route.mode === 'joint'
+        ? fuseAnswer(question, pairs[index], answers[name], scores, route)
+        : soundAnswer(question, pairs[index], scores);
+    }
+    clapMs = performance.now() - clapStarted;
+  }
+
+  return {model: 'mocovoice-audio-jev-edge-qwen-noulmatch-clap-learned-router-v6', execution: {...status(), audio_sent_to_server: false}, answers,
+    evidence: {audio_duration_ms: samples.length / 16, qwen_windows: qwenCount, clap_windows: clapWindows,
+      shared_audio_encoders: ['Qwen/Qwen3-ASR-0.6B', 'laion/clap-htsat-unfused'], audio_reused_across_questions: true,
+      question_routing: Object.fromEntries(entries.map(([name], index) => [name, publicRoute(routes[index])]))},
+    latency_ms: {qwen: qwenMs, clap: clapMs, total: performance.now() - started}, warnings: [...warnings],
+    scope: '研究用試作。未知タスクへの正確さと確率校正は未検証。'};
+}
+
+let queue = Promise.resolve();
+onmessage = ({data}) => {
+  queue = queue.then(async () => {
+    currentId = data.id;
+    try {
+      const result = data.type === 'load' ? await load(data.backend) : await analyze(data.samples, data.samples48k, data.context, data.questions);
+      postMessage({id: data.id, result});
+    } catch (error) {postMessage({id: data.id, error: error.message || String(error)});}
+  });
+};

inference.js

@@ -0,0 +1,34 @@
+// Only model assets leave/enter the network. User inputs go to a local Worker.
+let worker;
+let serial = 0;
+const pending = new Map();
+function client() {
+  if (!worker) {
+    worker = new Worker(new URL('./inference-worker.js?v=20260921-preload-all-v1', import.meta.url), {type: 'module'});
+    worker.onmessage = ({data}) => {
+      const call = pending.get(data.id);
+      if (!call) return;
+      if (data.progress) { call.onProgress?.(data.progress); return; }
+      pending.delete(data.id);
+      data.error ? call.reject(new Error(data.error)) : call.resolve(data.result);
+    };
+    worker.onerror = event => {
+      for (const call of pending.values()) call.reject(new Error(event.message || 'ブラウザ内の推論処理に失敗しました。'));
+      pending.clear(); worker.terminate(); worker = null;
+    };
+  }
+  return worker;
+}
+function request(type, payload, onProgress) {
+  const id = ++serial;
+  return new Promise((resolve, reject) => {
+    pending.set(id, {resolve, reject, onProgress});
+    client().postMessage({id, type, ...payload});
+  });
+}
+export function loadModel({onProgress, backend = 'auto'} = {}) {
+  return request('load', {backend}, onProgress);
+}
+export function analyze(samples16k, samples48k, context, questions, {onProgress} = {}) {
+  return request('analyze', {samples: samples16k, samples48k, context, questions}, onProgress);
+}

licenses/LINE-Seed-OFL.txt

@@ -0,0 +1,93 @@
+Copyright 2020-2022 LY Corporation
+
+This Font Software is licensed under the SIL Open Font License, Version 1.1.
+This license is copied below, and is also available with a FAQ at:
+http://scripts.sil.org/OFL
+
+
+-----------------------------------------------------------
+SIL OPEN FONT LICENSE Version 1.1 - 26 February 2007
+-----------------------------------------------------------
+
+PREAMBLE
+The goals of the Open Font License (OFL) are to stimulate worldwide
+development of collaborative font projects, to support the font creation
+efforts of academic and linguistic communities, and to provide a free and
+open framework in which fonts may be shared and improved in partnership
+with others.
+
+The OFL allows the licensed fonts to be used, studied, modified and
+redistributed freely as long as they are not sold by themselves. The
+fonts, including any derivative works, can be bundled, embedded, 
+redistributed and/or sold with any software provided that any reserved
+names are not used by derivative works. The fonts and derivatives,
+however, cannot be released under any other type of license. The
+requirement for fonts to remain under this license does not apply
+to any document created using the fonts or their derivatives.
+
+DEFINITIONS
+"Font Software" refers to the set of files released by the Copyright
+Holder(s) under this license and clearly marked as such. This may
+include source files, build scripts and documentation.
+
+"Reserved Font Name" refers to any names specified as such after the
+copyright statement(s).
+
+"Original Version" refers to the collection of Font Software components as
+distributed by the Copyright Holder(s).
+
+"Modified Version" refers to any derivative made by adding to, deleting,
+or substituting -- in part or in whole -- any of the components of the
+Original Version, by changing formats or by porting the Font Software to a
+new environment.
+
+"Author" refers to any designer, engineer, programmer, technical
+writer or other person who contributed to the Font Software.
+
+PERMISSION & CONDITIONS
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of the Font Software, to use, study, copy, merge, embed, modify,
+redistribute, and sell modified and unmodified copies of the Font
+Software, subject to the following conditions:
+
+1) Neither the Font Software nor any of its individual components,
+in Original or Modified Versions, may be sold by itself.
+
+2) Original or Modified Versions of the Font Software may be bundled,
+redistributed and/or sold with any software, provided that each copy
+contains the above copyright notice and this license. These can be
+included either as stand-alone text files, human-readable headers or
+in the appropriate machine-readable metadata fields within text or
+binary files as long as those fields can be easily viewed by the user.
+
+3) No Modified Version of the Font Software may use the Reserved Font
+Name(s) unless explicit written permission is granted by the corresponding
+Copyright Holder. This restriction only applies to the primary font name as
+presented to the users.
+
+4) The name(s) of the Copyright Holder(s) or the Author(s) of the Font
+Software shall not be used to promote, endorse or advertise any
+Modified Version, except to acknowledge the contribution(s) of the
+Copyright Holder(s) and the Author(s) or with their explicit written
+permission.
+
+5) The Font Software, modified or unmodified, in part or in whole,
+must be distributed entirely under this license, and must not be
+distributed under any other license. The requirement for fonts to
+remain under this license does not apply to any document created
+using the Font Software.
+
+TERMINATION
+This license becomes null and void if any of the above conditions are
+not met.
+
+DISCLAIMER
+THE FONT SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT
+OF COPYRIGHT, PATENT, TRADEMARK, OR OTHER RIGHT. IN NO EVENT SHALL THE
+COPYRIGHT HOLDER BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+INCLUDING ANY GENERAL, SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL
+DAMAGES, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+FROM, OUT OF THE USE OR INABILITY TO USE THE FONT SOFTWARE OR FROM
+OTHER DEALINGS IN THE FONT SOFTWARE.

licenses/clap-Apache-2.0.txt

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

licenses/clap-NOTICE.txt

@@ -0,0 +1,13 @@
+CLAP runtime notice
+
+The Cloud non-verbal branch uses laion/clap-htsat-unfused at revision
+8fa0f1c6d0433df6e97c127f64b2a1d6c0dcda8a.  Its model distribution is
+Apache-2.0 according to the pinned model card.
+
+Source: https://huggingface.co/laion/clap-htsat-unfused/tree/8fa0f1c6d0433df6e97c127f64b2a1d6c0dcda8a
+License: https://www.apache.org/licenses/LICENSE-2.0
+
+The bundled supervised heads are project artifacts:
+- clap_sound_v1.pt: broad labels animal, nature, music, environment
+- clap_animal8_oof_v1.pt: dog, cat, bird, insect, chicken_and_rooster, crow,
+  frog, cricket

licenses/hubert-Apache-2.0.txt

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

licenses/irodori-MIT.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Aratako
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

licenses/irodori-model-card.md

@@ -0,0 +1,132 @@
+---
+license: mit
+language:
+- ja
+pipeline_tag: text-to-speech
+tags:
+- speech
+- voice
+- tts
+base_model:
+- Aratako/Irodori-TTS-500M-v2
+---
+
+# Irodori-TTS-500M-v3
+
+[![Code](https://img.shields.io/badge/Code-GitHub-black)](https://github.com/Aratako/Irodori-TTS) [![WandB](https://img.shields.io/badge/Training%20Log-WandB-orange)](https://api.wandb.ai/links/aratako-lm/2ctrvcim) [![Demo Space](https://img.shields.io/badge/Demo-HuggingFace%20Space-blue)](https://huggingface.co/spaces/Aratako/Irodori-TTS-500M-v3-Demo)
+
+**Irodori-TTS-500M-v3** is a Japanese Text-to-Speech model based on a Rectified Flow Diffusion Transformer (RF-DiT) architecture. The architecture and training design largely follow [Echo-TTS](https://jordandarefsky.com/blog/2025/echo/), using continuous latents as the generation target. It supports zero-shot voice cloning from reference audio.
+
+A unique feature of this model is **emoji-based style and sound effect control** — by inserting specific emojis into the input text, you can control speaking styles, emotions, and even sound effects in the generated audio.
+
+## 🌟 Key Features
+
+  * **Flow Matching TTS:** Rectified Flow Diffusion Transformer over continuous DACVAE latents for high-quality Japanese speech synthesis.
+  * **Voice Cloning:** Zero-shot voice cloning from a short reference audio clip.
+  * **Emoji-based Style Control:** Control speaking styles, emotions, and sound effects by embedding emojis directly in the input text. See  [`EMOJI_ANNOTATIONS.md`](EMOJI_ANNOTATIONS.md) for the full list of supported emojis and their effects.
+
+## ✨ What's New in v3
+
+This version brings several major improvements over the previous v2 model:
+
+  * **Variable-length Training & Duration Predictor:** Transitioned from fixed-length to variable-length training and introduced a Duration Predictor. This improves training efficiency and enhances the Real-Time Factor (RTF) during inference.
+  * **Expanded Training Data:** The model was trained on a larger dataset, resulting in more natural speech synthesis and improved robustness across various speaking styles.
+  * **Integrated Watermarking:** Integrated [SilentCipher](https://github.com/sony/silentcipher) to apply robust, invisible audio watermarks directly to the generated outputs, promoting responsible AI usage.
+
+## 🏗️ Architecture
+
+The model (approximately 500M parameters) consists of four main components:
+
+1.  **Text Encoder:** Token embeddings initialized from [llm-jp/llm-jp-3-150m](https://huggingface.co/llm-jp/llm-jp-3-150m), followed by self-attention + SwiGLU transformer layers with RoPE.
+2.  **Reference Latent Encoder:** Encodes patched reference audio latents for speaker/style conditioning via self-attention + SwiGLU layers.
+3.  **Diffusion Transformer:** Joint-attention DiT blocks with Low-Rank AdaLN (timestep-conditioned adaptive layer normalization), half-RoPE, and SwiGLU MLPs.
+4.  **Duration Predictor:** Predicts audio duration from encoded text and speaker conditioning using stacked SwiGLU MLP blocks.
+
+Audio is represented as continuous latent sequences via the [Aratako/Semantic-DACVAE-Japanese-32dim](https://huggingface.co/Aratako/Semantic-DACVAE-Japanese-32dim) codec (32-dim), enabling high-quality 48kHz waveform reconstruction.
+
+## 🎧 Audio Samples
+
+### 1\. Standard TTS
+
+Basic Japanese text-to-speech generation (without reference audio).
+
+| Case | Text | Generated Audio |
+| :--- | :--- | :--- |
+| **Sample 1** | "お電話ありがとうございます。ただいま電話が大変混み合っております。恐れ入りますが、発信音のあとに、ご用件をお話しください。" | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/standard_sample1.wav"></audio> |
+| **Sample 2** | "その森には、古い言い伝えがありました。月が最も高く昇る夜、静かに耳を澄ませば、風の歌声が聞こえるというのです。私は半信半疑でしたが、その夜、確かに誰かが私を呼ぶ声を聞いたのです。" | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/standard_sample2.wav"></audio> |
+
+### 2\. Emoji Annotation Control
+
+Examples of controlling speaking style and effects with emojis. For the full list of supported emojis, see  [`EMOJI_ANNOTATIONS.md`](EMOJI_ANNOTATIONS.md).
+
+| Case | Text (with Emoji) | Generated Audio |
+| :--- | :--- | :--- |
+| **Sample 1** | なーに、どうしたの？…え？もっと近づいてほしい？…👂😮‍💨👂😮‍💨こういうのが好きなんだ？ | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/emoji_sample1.wav"></audio> |
+| **Sample 2** | うぅ…😭そんなに酷いこと、言わないで…😭 | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/emoji_sample2.wav"></audio> |
+| **Sample 3** | 🤧🤧ごめんね、風邪引いちゃってて🤧…大丈夫、ただの風邪だ��らすぐ治るよ🥺 | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/emoji_sample3.wav"></audio> |
+
+### 3\. Voice Cloning (Zero-shot)
+
+Examples of cloning a voice from a reference audio clip.
+
+| Case | Reference Audio | Generated Audio |
+| :--- | :--- | :--- |
+| **Example 1** | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/clone_ref1.wav"></audio> | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/clone_gen1.wav"></audio> |
+| **Example 2** | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/clone_ref2.wav"></audio> | <audio controls src="https://huggingface.co/Aratako/Irodori-TTS-500M-v3/resolve/main/samples/clone_gen2.wav"></audio> |
+
+## 🚀 Usage
+
+For inference code, installation instructions, and training scripts, please refer to the GitHub repository:
+
+👉 **[GitHub: Aratako/Irodori-TTS](https://github.com/Aratako/Irodori-TTS)**
+
+## 📊 Training Data & Annotation
+
+The model was trained on an expanded, high-quality Japanese speech dataset. To enable the emoji-based style control, the training texts were enriched with emoji annotations. These annotations were automatically generated and labeled using a fine-tuned model based on [Qwen/Qwen3-Omni-30B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct).
+
+## ⚠️ Limitations
+
+  - **Japanese Only:** This model currently supports Japanese text input only.
+  - **Emoji Control:** While emoji-based style control adds expressiveness, the effect may vary depending on context and is not always perfectly consistent.
+  - **Audio Quality:** Quality depends on training data characteristics. Performance may vary for voices or speaking styles underrepresented in the training data.
+  - **Kanji Reading Accuracy:** The model's ability to accurately read Kanji is relatively weak compared to other TTS models of a similar size. You may need to convert complex Kanji into Hiragana or Katakana beforehand.
+
+## 📜 License & Ethical Restrictions
+
+### License
+
+This model is released under **[MIT](https://choosealicense.com/licenses/mit/)**.
+
+### Ethical Restrictions
+
+In addition to the license terms, the following ethical restrictions apply:
+
+1.  **No Impersonation:** Do not use this model to clone or impersonate the voice of any individual (e.g., voice actors, celebrities, public figures) without their explicit consent.
+2.  **No Misinformation:** Do not use this model to generate deepfakes or synthetic speech intended to mislead others or spread misinformation.
+3.  **Voice Generation Disclaimer:** When generating speech purely from text without using a reference audio, it is possible that the generated voice may coincidentally resemble that of a real person. This is strictly a probabilistic artifact within the latent space. The model was not trained with the intent of reproducing specific individuals.
+4.  **Liability Disclaimer:** The developers assume no liability for any misuse of this model. Users are solely responsible for ensuring their use of the generated content complies with applicable laws and regulations in their jurisdiction.
+
+## 🙏 Acknowledgments
+
+This project builds upon the following works:
+
+  - [Echo-TTS](https://jordandarefsky.com/blog/2025/echo/) — Architecture and training design reference
+  - [DACVAE](https://github.com/facebookresearch/dacvae) — Audio VAE
+  - [llm-jp/llm-jp-3-150m](https://huggingface.co/llm-jp/llm-jp-3-150m) — Tokenizer and embedding weight initialization
+  - [SilentCipher](https://github.com/sony/silentcipher) — Audio watermarking integration
+
+We would also like to extend our special thanks to **[Respair](https://huggingface.co/Respair)** for the inspiration behind the emoji annotation feature, and to [gabrielclark3330](https://huggingface.co/gabrielclark3330) for supporting this project.
+
+## 🖊️ Citation
+
+If you use Irodori-TTS-v3 in your research or project, please cite it as follows:
+
+```bibtex
+@misc{irodori-tts-v3,
+  author = {Chihiro Arata},
+  title = {Irodori-TTS: A Flow Matching-based Text-to-Speech Model with Emoji-driven Style Control},
+  year = {2026},
+  publisher = {Hugging Face},
+  journal = {Hugging Face repository},
+  howpublished = {\url{https://huggingface.co/Aratako/Irodori-TTS-500M-v3}}
+}

licenses/jszip.txt

@@ -0,0 +1,651 @@
+JSZip is dual licensed. At your choice you may use it under the MIT license *or* the GPLv3
+license.
+
+The MIT License
+===============
+
+Copyright (c) 2009-2016 Stuart Knightley, David Duponchel, Franz Buchinger, António Afonso
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+
+
+GPL version 3
+=============
+
+                    GNU GENERAL PUBLIC LICENSE
+                       Version 3, 29 June 2007
+
+ Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                            Preamble
+
+  The GNU General Public License is a free, copyleft license for
+software and other kinds of works.
+
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+the GNU General Public License is intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.  We, the Free Software Foundation, use the
+GNU General Public License for most of our software; it applies also to
+any other work released this way by its authors.  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
+
+  To protect your rights, we need to prevent others from denying you
+these rights or asking you to surrender the rights.  Therefore, you have
+certain responsibilities if you distribute copies of the software, or if
+you modify it: responsibilities to respect the freedom of others.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must pass on to the recipients the same
+freedoms that you received.  You must make sure that they, too, receive
+or can get the source code.  And you must show them these terms so they
+know their rights.
+
+  Developers that use the GNU GPL protect your rights with two steps:
+(1) assert copyright on the software, and (2) offer you this License
+giving you legal permission to copy, distribute and/or modify it.
+
+  For the developers' and authors' protection, the GPL clearly explains
+that there is no warranty for this free software.  For both users' and
+authors' sake, the GPL requires that modified versions be marked as
+changed, so that their problems will not be attributed erroneously to
+authors of previous versions.
+
+  Some devices are designed to deny users access to install or run
+modified versions of the software inside them, although the manufacturer
+can do so.  This is fundamentally incompatible with the aim of
+protecting users' freedom to change the software.  The systematic
+pattern of such abuse occurs in the area of products for individuals to
+use, which is precisely where it is most unacceptable.  Therefore, we
+have designed this version of the GPL to prohibit the practice for those
+products.  If such problems arise substantially in other domains, we
+stand ready to extend this provision to those domains in future versions
+of the GPL, as needed to protect the freedom of users.
+
+  Finally, every program is threatened constantly by software patents.
+States should not allow patents to restrict development and use of
+software on general-purpose computers, but in those that do, we wish to
+avoid the special danger that patents applied to a free program could
+make it effectively proprietary.  To prevent this, the GPL assures that
+patents cannot be used to render the program non-free.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                       TERMS AND CONDITIONS
+
+  0. Definitions.
+
+  "This License" refers to version 3 of the GNU General Public License.
+
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
+
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
+
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
+
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Use with the GNU Affero General Public License.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU Affero General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the special requirements of the GNU Affero General Public License,
+section 13, concerning interaction through a network will apply to the
+combination as such.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS