Delete app.py

app.py

@@ -1,215 +0,0 @@
-from fastapi import FastAPI, File, UploadFile, HTTPException, Form
-from fastapi.responses import PlainTextResponse
-import whisper
-import tempfile
-import os
-import sys
-import json
-import time
-from pathlib import Path
-from typing import Optional
-
-# 把 stderr 合并到 stdout，HuggingFace Spaces log 才能显示 Whisper 进度条
-sys.stderr = sys.stdout
-
-app = FastAPI(title="Subtitle Generator API")
-
-model_size = os.getenv("WHISPER_MODEL", "medium")
-print(f"[Startup] 加载 Whisper 模型: {model_size} ...", flush=True)
-model = whisper.load_model(model_size)
-print(f"[Startup] 模型加载完成 ✅", flush=True)
-
-
-import re
-
-# 每种语言的默认分段字数（用户未显式传 max_chars 时生效）
-DEFAULT_MAX_CHARS = {"zh": 12, "en": 55}
-FALLBACK_MAX_CHARS = 40  # 其他语言兜底值
-
-# 需要从最终字幕文本中清除的标点符号（中英文标点、引号、括号等）
-# 特意不包含直引号撇号 ' ，避免破坏英文缩写（don't / it's）
-_PUNCT_CHARS = "，。！？、；：“”‘’「」『』（）()《》【】[]<>—–…‥·,.!?;:\""
-_PUNCT_PATTERN = re.compile("[" + re.escape(_PUNCT_CHARS) + "]")
-
-
-def strip_punctuation(text: str) -> str:
-    """去除字幕文本中的标点符号，仅做视觉清洗，不影响时间戳/断句逻辑"""
-    if not text:
-        return text
-    cleaned = _PUNCT_PATTERN.sub("", text)
-    cleaned = re.sub(r"[ \t]+", " ", cleaned).strip()
-    return cleaned
-
-
-def resolve_max_chars(user_max_chars: int, lang_code: str) -> int:
-    """决定实际使用的 max_chars：用户显式指定则用用户的，否则按语言给默认值"""
-    if user_max_chars and user_max_chars > 0:
-        return user_max_chars
-    base_lang = (lang_code or "").split("-")[0].lower()
-    return DEFAULT_MAX_CHARS.get(base_lang, FALLBACK_MAX_CHARS)
-
-
-def format_timestamp(seconds: float) -> str:
-    hours = int(seconds // 3600)
-    minutes = int((seconds % 3600) // 60)
-    secs = int(seconds % 60)
-    millisecs = int((seconds - int(seconds)) * 1000)
-    return f"{hours:02d}:{minutes:02d}:{secs:02d},{millisecs:03d}"
-
-
-def group_by_chars(result, max_chars: int):
-    """用词级时间戳按字数重新组合字幕段，时间戳精确来自 Whisper"""
-    words = []
-    for seg in result["segments"]:
-        for w in seg.get("words", []):
-            word = w["word"].strip()
-            if word:
-                words.append({"word": word, "start": w["start"], "end": w["end"]})
-
-    if not words:
-        print("[Warning] 没有词级时间戳，回退到原始分段", flush=True)
-        return result["segments"]
-
-    segments = []
-    buf_text = ""
-    buf_start = None
-    buf_end = None
-
-    for w in words:
-        if buf_start is None:
-            buf_start = w["start"]
-        buf_text += w["word"]
-        buf_end = w["end"]
-
-        if len(buf_text) >= max_chars or buf_text[-1] in "，。！？、,.!?":
-            segments.append({"start": buf_start, "end": buf_end, "text": buf_text})
-            buf_text = ""
-            buf_start = None
-
-    if buf_text:
-        segments.append({"start": buf_start, "end": buf_end, "text": buf_text})
-
-    # 断句依赖标点判断已经用完了，这里统一清洗掉标点，不影响时间戳
-    for seg in segments:
-        seg["text"] = strip_punctuation(seg["text"])
-
-    return segments
-
-
-def generate_srt(segments) -> str:
-    srt_content = []
-    for i, segment in enumerate(segments, 1):
-        start_time = format_timestamp(segment['start'])
-        end_time = format_timestamp(segment['end'])
-        text = segment['text'].strip()
-        srt_content.append(f"{i}\n{start_time} --> {end_time}\n{text}\n")
-    return "\n".join(srt_content)
-
-
-def generate_vtt(segments) -> str:
-    vtt_content = ["WEBVTT\n"]
-    for i, segment in enumerate(segments):
-        start_time = format_timestamp(segment['start']).replace(',', '.')
-        end_time = format_timestamp(segment['end']).replace(',', '.')
-        text = segment['text'].strip()
-        vtt_content.append(f"{start_time} --> {end_time}\n{text}\n")
-    return "\n".join(vtt_content)
-
-
-@app.get("/")
-async def root():
-    return {
-        "message": "Subtitle Generator API",
-        "model": model_size,
-        "device": "cpu"
-    }
-
-
-@app.get("/health")
-async def health_check():
-    return {"status": "healthy", "model": model_size}
-
-
-@app.post("/generate_subtitle")
-async def generate_subtitle(
-    file: UploadFile = File(...),
-    output_format: str = Form("srt"),
-    language: Optional[str] = Form(None),
-    max_chars: int = Form(0)  # 0/未传 = 按语言自动使用默认值（zh=12, en=55, 其他=40）
-):
-    valid_formats = ["srt", "vtt", "txt", "json"]
-    if output_format not in valid_formats:
-        raise HTTPException(status_code=400, detail=f"Invalid format. Choose from: {valid_formats}")
-
-    suffix = Path(file.filename).suffix if file.filename else ".tmp"
-    with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp_file:
-        content = await file.read()
-        tmp_file.write(content)
-        tmp_path = tmp_file.name
-
-    file_mb = len(content) / 1024 / 1024
-    print(f"[Transcribe] 文件: {file.filename} ({file_mb:.1f} MB)", flush=True)
-    print(f"[Transcribe] 格式: {output_format} | 语言: {language or '自动检测'} | max_chars: {max_chars if max_chars > 0 else '未指定(将按语言自动选择)'}", flush=True)
-
-    try:
-        # word_timestamps 永远开启：分组逻辑现在始终执行，不再有"完全不分组"的原始模式
-        transcribe_options = {"verbose": True, "word_timestamps": True}
-        if language:
-            transcribe_options["language"] = language
-
-        print(f"[Transcribe] 开始转写，请稍候...", flush=True)
-        t0 = time.time()
-        result = model.transcribe(tmp_path, **transcribe_options)
-        elapsed = time.time() - t0
-
-        seg_count = len(result["segments"])
-        detected_lang = result.get("language", "unknown")
-        print(f"[Transcribe] 完成 ✅ 耗时 {elapsed:.1f}s | 语言: {detected_lang} | 原始段数: {seg_count}", flush=True)
-
-        effective_max_chars = resolve_max_chars(max_chars, language or detected_lang)
-        source = "用户指定" if (max_chars and max_chars > 0) else f"按语言({language or detected_lang})自动默认"
-        segments = group_by_chars(result, effective_max_chars)
-        print(f"[Transcribe] 按字数拆分完成，共 {len(segments)} 段（max_chars={effective_max_chars}，{source}）", flush=True)
-
-        full_text = strip_punctuation(result["text"])
-
-        if output_format == "srt":
-            subtitle_text = generate_srt(segments)
-            media_type = "text/plain"
-            filename = "subtitle.srt"
-        elif output_format == "vtt":
-            subtitle_text = generate_vtt(segments)
-            media_type = "text/plain"
-            filename = "subtitle.vtt"
-        elif output_format == "txt":
-            subtitle_text = full_text
-            media_type = "text/plain"
-            filename = "subtitle.txt"
-        elif output_format == "json":
-            subtitle_text = json.dumps({
-                "text": full_text,
-                "segments": segments,
-                "language": result["language"],
-                "max_chars_used": effective_max_chars
-            }, ensure_ascii=False, indent=2)
-            media_type = "application/json"
-            filename = "subtitle.json"
-
-        print(f"[Transcribe] 返回结果: {filename}", flush=True)
-        return PlainTextResponse(
-            content=subtitle_text,
-            media_type=media_type,
-            headers={"Content-Disposition": f"attachment; filename={filename}"}
-        )
-
-    except Exception as e:
-        print(f"[Error] 转写失败: {e}", flush=True)
-        raise HTTPException(status_code=500, detail=f"Transcription failed: {str(e)}")
-    finally:
-        if os.path.exists(tmp_path):
-            os.unlink(tmp_path)
-
-
-if __name__ == "__main__":
-    import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=7860)