Upload 3 files

Dockerfile

@@ -0,0 +1,12 @@
+FROM pytorch/pytorch:2.1.0-cuda11.8-cudnn8-runtime
+
+WORKDIR /app
+
+RUN apt-get update && apt-get install -y ffmpeg git && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY . .
+
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -0,0 +1,118 @@
+from fastapi import FastAPI, UploadFile, File, Form
+import os, uuid, subprocess, torch, cv2
+import whisper
+from scenedetect import VideoManager, SceneManager
+from scenedetect.detectors import ContentDetector
+from ultralytics import YOLO
+from diffusers import StableVideoDiffusionPipeline
+
+app = FastAPI()
+
+UPLOAD_DIR = "uploads"
+OUTPUT_DIR = "outputs"
+os.makedirs(UPLOAD_DIR, exist_ok=True)
+os.makedirs(OUTPUT_DIR, exist_ok=True)
+
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+
+# ===== Load models =====
+
+whisper_model = whisper.load_model("base")
+
+yolo = YOLO("yolov8n.pt")
+
+svd = StableVideoDiffusionPipeline.from_pretrained(
+    "stabilityai/stable-video-diffusion-img2vid",
+    torch_dtype=torch.float16
+).to(DEVICE)
+
+# ===== Endpoints =====
+
+@app.post("/captions")
+async def captions(file: UploadFile = File(...)):
+    path = os.path.join(UPLOAD_DIR, file.filename)
+    with open(path, "wb") as f:
+        f.write(await file.read())
+
+    result = whisper_model.transcribe(path)
+
+    return {
+        "segments": result["segments"],
+        "language": result["language"]
+    }
+
+
+@app.post("/scene-detect")
+async def scene_detect(file: UploadFile = File(...)):
+    path = os.path.join(UPLOAD_DIR, file.filename)
+    with open(path, "wb") as f:
+        f.write(await file.read())
+
+    video_manager = VideoManager([path])
+    scene_manager = SceneManager()
+    scene_manager.add_detector(ContentDetector(threshold=27.0))
+
+    video_manager.start()
+    scene_manager.detect_scenes(frame_source=video_manager)
+
+    scenes = scene_manager.get_scene_list()
+    video_manager.release()
+
+    return {
+        "scenes": [
+            {"start": s[0].get_seconds(), "end": s[1].get_seconds()}
+            for s in scenes
+        ]
+    }
+
+
+@app.post("/smart-crop")
+async def smart_crop(
+    file: UploadFile = File(...),
+    aspect: str = Form("9:16")
+):
+    path = os.path.join(UPLOAD_DIR, file.filename)
+    with open(path, "wb") as f:
+        f.write(await file.read())
+
+    cap = cv2.VideoCapture(path)
+    ret, frame = cap.read()
+    cap.release()
+
+    results = yolo(frame)
+    box = results[0].boxes.xyxy[0].cpu().numpy()
+
+    return {
+        "crop_box": box.tolist(),
+        "aspect": aspect
+    }
+
+
+@app.post("/edit")
+async def edit_video(
+    file: UploadFile = File(...),
+    prompt: str = Form(...)
+):
+    path = os.path.join(UPLOAD_DIR, file.filename)
+    with open(path, "wb") as f:
+        f.write(await file.read())
+
+    # Extract first frame
+    subprocess.run([
+        "ffmpeg", "-i", path, "-frames:v", "1",
+        "frame.png"
+    ])
+
+    from PIL import Image
+    img = Image.open("frame.png").resize((512, 512))
+
+    frames = svd(
+        image=img,
+        prompt=prompt,
+        num_frames=16
+    ).frames
+
+    return {
+        "prompt": prompt,
+        "frames_generated": len(frames)
+    }

requirements.txt

@@ -0,0 +1,17 @@
+fastapi
+uvicorn
+torch
+torchaudio
+torchvision
+opencv-python
+numpy
+scipy
+ffmpeg-python
+
+openai-whisper
+pyscenedetect
+ultralytics
+
+diffusers
+transformers
+accelerate