Create modules/lut.py

.gitattributes

@@ -0,0 +1,35 @@
+*.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

Dockerfile

@@ -1,37 +1,6 @@
-FROM ubuntu:24.04
-
-# acestep.cpp builds from source with cmake. CPU-only path (matches
-# buildcpu.sh: cmake -DGGML_BLAS=ON). No CUDA/Vulkan/Metal.
-# pkg-config is REQUIRED: ggml-blas/CMakeLists.txt calls find_package(PkgConfig)
-# and fails the configure step without it (build died here on the first attempt).
-RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
-    git cmake build-essential ca-certificates curl python3 \
-    pkg-config \
-    libblas-dev liblapack-dev libgomp1 libcurl4 \
-    && rm -rf /var/lib/apt/lists/*
-
-WORKDIR /build
-
-# clone source + ggml submodule
-RUN git clone --depth 1 --recurse-submodules \
-      https://github.com/ServeurpersoCom/acestep.cpp.git . \
-    && rm -rf build && mkdir build && cd build \
-    && cmake .. -DGGML_BLAS=ON -DCMAKE_BUILD_TYPE=Release \
-    && cmake --build . --config Release -j "$(getconf _NPROCESSORS_ONLN)"
-
-# weights: 0.6B LM + Q8_0 DiT turbo + Q8_0 text encoder + BF16 VAE (~4.2 GB total)
-RUN mkdir -p /app/models \
-    && curl -fL -o /app/models/acestep-5Hz-lm-0.6B-Q8_0.gguf \
-      https://huggingface.co/Serveurperso/ACE-Step-1.5-GGUF/resolve/main/acestep-5Hz-lm-0.6B-Q8_0.gguf \
-    && curl -fL -o /app/models/acestep-v15-turbo-Q8_0.gguf \
-      https://huggingface.co/Serveurperso/ACE-Step-1.5-GGUF/resolve/main/acestep-v15-turbo-Q8_0.gguf \
-    && curl -fL -o /app/models/Qwen3-Embedding-0.6B-Q8_0.gguf \
-      https://huggingface.co/Serveurperso/ACE-Step-1.5-GGUF/resolve/main/Qwen3-Embedding-0.6B-Q8_0.gguf \
-    && curl -fL -o /app/models/vae-BF16.gguf \
-      https://huggingface.co/Serveurperso/ACE-Step-1.5-GGUF/resolve/main/vae-BF16.gguf
-
+FROM python:3.10-slim
+RUN apt-get update && apt-get install -y ffmpeg fonts-noto-cjk
+COPY . /app
 WORKDIR /app
-COPY run.sh /app/run.sh
-RUN chmod +x /app/run.sh
-EXPOSE 7860
-CMD ["bash", "/app/run.sh"]
+RUN pip install -r requirements.txt
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,88 +1,10 @@
 ---
-title: ACE-Step Music
-emoji: 🎵
-colorFrom: purple
-colorTo: pink
+title: Video Gen
+emoji: 📊
+colorFrom: green
+colorTo: red
 sdk: docker
-app_port: 7860
 pinned: false
 ---
 
-# 🎵 ACE-Step Music (acestep.cpp, CPU)
-
-AI music generation on cpu-basic. Text + lyrics in, stereo 48kHz audio out.
-Built from source (`acestep.cpp` GGML) with the **0.6B** model set:
-
-| 部件 | 文件 | 大小 |
-|---|---|---|
-| LM（歌词+音频编码） | `acestep-5Hz-lm-0.6B-Q8_0.gguf` | 0.68 GB |
-| DiT（flow-matching 扩散，turbo 8 步） | `acestep-v15-turbo-Q8_0.gguf` | 2.4 GB |
-| 文本编码器 | `Qwen3-Embedding-0.6B-Q8_0.gguf` | 0.75 GB |
-| VAE | `vae-BF16.gguf` | 0.32 GB |
-
-合计约 **4.2 GB**，16 GB 内存装得下。占用会随生成过程浮动，峰值约 6 GB。
-
-## 使用
-
-网页 UI 在 `/`（`ace-server` 内嵌的 WebUI，无需 npm）。写 caption → 设时长 → 生成 → 下载。
-
-## REST API（给 pipeline 调用）
-
-异步作业模型：POST 拿 id → 轮询 `/job?id=` → `result=1` 取文件。
-
-**请求体（`examples/dit-only.json` 格式）：**
-```json
-{
-  "lm_model": "acestep-5Hz-lm-0.6B-Q8_0.gguf",
-  "synth_model": "acestep-v15-turbo-Q8_0.gguf",
-  "caption": "Ambient electronic soundscape with warm analog pads",
-  "lyrics": "",
-  "bpm": 90,
-  "duration": 30,
-  "keyscale": "C minor",
-  "timesignature": "4",
-  "vocal_language": "en",
-  "inference_steps": 8,
-  "guidance_scale": 1.0,
-  "shift": 3.0
-}
-```
-
-### 出音乐（两步）
-```bash
-# 1) LM: caption -> lyrics + audio codes
-LM_ID=$(curl -sX POST https://jacky3102-platform.hf.space/lm \
-  -H 'Content-Type: application/json' -d @req.json | jq -r '.id')
-
-# 2) 轮询直到 done
-curl -s "https://jacky3102-platform.hf.space/job?id=$LM_ID&result=1" | jq '.[0]' > out.json
-
-# 3) synth: codes -> audio (WAV or MP3)
-SY_ID=$(curl -sX POST https://jacky3102-platform.hf.space/synth \
-  -H 'Content-Type: application/json' -d @out.json | jq -r '.id')
-curl -s "https://jacky3102-platform.hf.space/job?id=$SY_ID&result=1" -o track.wav
-```
-
-### 可用模型 / 默认参数
-```bash
-curl https://jacky3102-platform.hf.space/props
-```
-
-### 跳过 LM（直接用已知歌词）
-POST 同一个 `/synth`，带 `lyrics` + `lm_model` 即可（dit-only 模式）。
-
-## 已知限制
-
-- **CPU 生成慢**：30 秒音乐在 2 vCPU 上预计几分钟到十几分钟，未实测。第一次请求会额外等模型加载（`--keep-loaded` 已开，之后不再重复加载）。
-- HF Space 有请求超时，长生成要用 `/job` 异步轮询，别在线等。
-- 官方 Space 用 ZeroGPU；这是 CPU 移植版，慢但免费。
-
-## 模型质量档位（之后可换）
-
-| LM | 大小 | 说明 |
-|---|---|---|
-| `acestep-5Hz-lm-0.6B-Q8_0.gguf` | 0.68 GB | 最快（当前） |
-| `acestep-5Hz-lm-1.7B-Q8_0.gguf` | 1.9 GB | 折中 |
-| `acestep-5Hz-lm-4B-Q8_0.gguf` | 4.2 GB | 官方默认，音质最好 |
-
-DiT 也可换 `sft`（50 步，音质更高但慢一倍）。
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,141 @@
+from fastapi import FastAPI, File, UploadFile, HTTPException, Form
+from fastapi.responses import Response
+import tempfile
+import os
+import traceback
+import sys
+from pathlib import Path
+from typing import Optional, List
+
+from modules.video_processor import generate_video
+from modules.lut import LUTS
+from modules.particles import EFFECTS
+from modules.titles import TITLE_STYLES, TITLE_ANIMATIONS, TITLE_POSITIONS
+
+app = FastAPI(title="Video Generator with Effects API")
+
+@app.get("/")
+async def root():
+    return {
+        "message": "Video Generator API",
+        "effects": EFFECTS,
+        "luts": {k: v["name"] for k, v in LUTS.items()},
+        "title_styles": TITLE_STYLES,
+        "title_animations": TITLE_ANIMATIONS,
+        "title_positions": TITLE_POSITIONS,
+        "features": ["ken_burns", "typewriter", "cinemascope", "title", "video_input"],
+    }
+
+@app.get("/health")
+async def health():
+    return {"status": "healthy"}
+
+@app.post("/generate")
+async def generate(
+    images: List[UploadFile] = File(None),
+    video: UploadFile = File(None),
+    audio: UploadFile = File(...),
+    srt_text: str = Form(""),
+    bgm: Optional[UploadFile] = File(None),
+    effect: str = Form("snow"),
+    lut: str = Form("none"),
+    ken_burns: bool = Form(False),
+    typewriter: bool = Form(False),
+    cinemascope: bool = Form(False),
+    title_text: str = Form(""),
+    title_style: str = Form("3d"),
+    title_color: str = Form("white"),
+    title_animation: str = Form("fade_in"),
+    title_duration: float = Form(4.0),
+    title_position: str = Form("center"),
+    title_start: float = Form(0.0),
+    title_font_size: int = Form(0),
+    typewriter_font_size: int = Form(0),
+    typewriter_color: str = Form("white"),
+    subtitle_color: str = Form("white"),
+    subtitle_effect: str = Form("none"),
+    fps: int = Form(25),
+):
+    # 验证
+    if not images and not video:
+        raise HTTPException(400, "必须提供 images 或 video 中的一个")
+    if images and video:
+        raise HTTPException(400, "不能同时提供 images 和 video")
+    
+    # 创建临时目录
+    tmp_dir = tempfile.mkdtemp()
+    try:
+        # 保存文件
+        image_paths = None
+        video_path = None
+        if images:
+            image_paths = []
+            for i, img in enumerate(images):
+                suffix = Path(img.filename).suffix if img.filename else ".jpg"
+                path = os.path.join(tmp_dir, f"image_{i}{suffix}")
+                with open(path, "wb") as f:
+                    f.write(await img.read())
+                image_paths.append(path)
+        elif video:
+            suffix = Path(video.filename).suffix if video.filename else ".mp4"
+            video_path = os.path.join(tmp_dir, f"video{suffix}")
+            with open(video_path, "wb") as f:
+                f.write(await video.read())
+
+        audio_path = os.path.join(tmp_dir, "audio" + (Path(audio.filename).suffix or ".wav"))
+        with open(audio_path, "wb") as f:
+            f.write(await audio.read())
+
+        bgm_path = None
+        if bgm:
+            bgm_path = os.path.join(tmp_dir, "bgm" + (Path(bgm.filename).suffix or ".mp3"))
+            with open(bgm_path, "wb") as f:
+                f.write(await bgm.read())
+
+        output_path = os.path.join(tmp_dir, "output.mp4")
+        
+        # 调用主合成函数
+        generate_video(
+            image_paths=image_paths,
+            video_path=video_path,
+            audio_path=audio_path,
+            srt_text=srt_text,
+            bgm_path=bgm_path,
+            effect=effect,
+            output_path=output_path,
+            lut=lut,
+            ken_burns=ken_burns,
+            typewriter=typewriter,
+            cinemascope=cinemascope,
+            title_text=title_text,
+            title_style=title_style,
+            title_color=title_color,
+            title_animation=title_animation,
+            title_duration=title_duration,
+            title_position=title_position,
+            title_start=title_start,
+            title_font_size=title_font_size,
+            typewriter_font_size=typewriter_font_size,
+            typewriter_color=typewriter_color,
+            subtitle_color=subtitle_color,
+            subtitle_effect=subtitle_effect,
+            fps=fps,
+        )
+
+        with open(output_path, "rb") as f:
+            video_bytes = f.read()
+
+        import shutil
+        shutil.rmtree(tmp_dir, ignore_errors=True)
+        return Response(content=video_bytes, media_type="video/mp4",
+                        headers={"Content-Disposition": "attachment; filename=output.mp4"})
+
+    except Exception as e:
+        print(traceback.format_exc(), file=sys.stderr)
+        import shutil
+        shutil.rmtree(tmp_dir, ignore_errors=True)
+        raise HTTPException(500, detail=str(e))
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)

modules/lut.py

@@ -0,0 +1,47 @@
+import numpy as np
+
+LUTS = {
+    "cinematic": {
+        "name": "🎬 电影橙青",
+        "matrix": np.array([[1.1, 0.0, -0.1], [0.0, 1.05, 0.0], [-0.05, 0.0, 1.15]]),
+        "warmth": (5, 0, -5), "contrast": 1.15,
+    },
+    "bw": {
+        "name": "⬛ 经典黑白",
+        "matrix": np.array([[0.3, 0.59, 0.11], [0.3, 0.59, 0.11], [0.3, 0.59, 0.11]]),
+        "warmth": (0, 0, 0), "contrast": 1.3,
+    },
+    "cyberpunk": {
+        "name": "🌆 赛博朋克",
+        "matrix": np.array([[0.8, 0.0, 0.3], [0.0, 0.9, 0.1], [0.1, 0.0, 1.3]]),
+        "warmth": (-10, 0, 15), "contrast": 1.2,
+    },
+    "warm": {
+        "name": "☀️ 日系暖阳",
+        "matrix": np.array([[1.1, 0.05, 0.0], [0.0, 1.0, -0.05], [-0.05, -0.05, 0.9]]),
+        "warmth": (15, 5, -10), "contrast": 0.95,
+    },
+    "noir": {
+        "name": "🌙 夜景蓝调",
+        "matrix": np.array([[0.7, 0.0, 0.2], [0.0, 0.8, 0.1], [0.1, 0.0, 1.2]]),
+        "warmth": (-15, -5, 10), "contrast": 1.25,
+    },
+}
+
+def apply_lut(frame, lut_name):
+    lut = LUTS.get(lut_name)
+    if not lut:
+        return frame
+    frame_float = frame.astype(np.float32)
+    h, w = frame.shape[:2]
+    pixels = frame_float.reshape(-1, 3)
+    transformed = pixels @ lut["matrix"].T
+    transformed = np.clip(transformed, 0, 255).reshape(h, w, 3)
+    b, g, r = cv2.split(transformed.astype(np.uint8))
+    r = cv2.add(r, lut["warmth"][0])
+    g = cv2.add(g, lut["warmth"][1])
+    b = cv2.add(b, lut["warmth"][2])
+    frame = cv2.merge([b, g, r])
+    frame = frame.astype(np.float32)
+    frame = (frame - 128) * lut["contrast"] + 128
+    return np.clip(frame, 0, 255).astype(np.uint8)

requirements.txt

@@ -0,0 +1,8 @@
+fastapi==0.104.1
+uvicorn==0.24.0
+python-multipart==0.0.6
+numpy>=1.24.0
+opencv-python-headless>=4.8.0
+Pillow>=10.0.0
+pydub>=0.25.0
+librosa>=0.10.0

run.sh

@@ -1,21 +0,0 @@
-#!/bin/bash
-# Boot diagnostics + ace-server on the port HF expects (7860).
-# On a crash the container stays alive so the traceback stays readable in the
-# runtime logs (this is what finally exposed the gradio/huggingface_hub bug
-# on Platform).
-echo "=== boot $(date -u) ==="
-echo "--- binaries ---"; ls /build/build/ | grep -E "^ace-" | head
-echo "--- models ---"; ls -la /app/models/
-echo "--- nproc / mem ---"; nproc; free -m | head -2
-echo "--- starting ace-server on 7860 ---"
-/build/build/ace-server \
-  --host 0.0.0.0 \
-  --port "${PORT:-7860}" \
-  --models /app/models \
-  --adapters /build/adapters \
-  --max-batch 1 \
-  --keep-loaded \
-  --device CPU
-echo "=== ace-server exited code=$? at $(date -u) ==="
-echo "--- holding container alive for log capture ---"
-sleep infinity