Clipper is a Python-based pipeline that takes a YouTube URL and produces TikTok-style vertical clips (9:16).
By default, it generates 8 clips of 45 seconds each.
Note: Clipper currently crops from the center of the video. Face-aware framing/motion tracking will be added in future releases.
- ✅ Input: YouTube URL
- ✅ Output: Multiple 9:16 vertical clips
- ✅ Defaults: 8 clips × 45 seconds
- ✅ Automatic temp & output folder creation
- ✅ Simple CLI interface
- OS: Windows / macOS / Linux
- Python: 3.9–3.12
- GPU: NVIDIA GPU recommended (CUDA) for best performance
- FFmpeg: Required and must be available on PATH
- PyTorch: Install with the CUDA build that matches your system (see below)
Windows (PowerShell):
python -m venv venv
.\venv\Scripts\activatemacOS/Linux (bash/zsh):
python -m venv venv
source venv/bin/activate- PyTorch + CUDA note (important) Before installing requirements.txt, ensure the PyTorch wheels match your CUDA runtime.
In requirements.txt, you will see entries like:
torch --index-url https://download.pytorch.org/whl/cu<Version>
torchvision --index-url https://download.pytorch.org/whl/cu<Version>
torchaudio --index-url https://download.pytorch.org/whl/cu<Version>Replace with the compatible CUDA build for your system, e.g.:
cu121 → CUDA 12.1
cu122 → CUDA 12.2
cu124 → CUDA 12.4
cu128 → CUDA 12.8
Example replacement:
torch --index-url https://download.pytorch.org/whl/cu128
torchvision --index-url https://download.pytorch.org/whl/cu128
torchaudio --index-url https://download.pytorch.org/whl/cu128If you don’t have an NVIDIA GPU or prefer CPU-only, use the CPU wheels from the official PyTorch site and remove the CUDA suffix.
-
Install dependencies
pip install --upgrade pip pip install -r requirements.txt
FFmpeg: If you don’t have FFmpeg, install it and make sure it’s on PATH.
Windows: winget install Gyan.FFmpeg or download a static build and add bin to PATH
macOS: brew install ffmpeg
Linux (Debian/Ubuntu):
sudo apt-get install ffmpeg
Copy the YouTube URL of the video you want to clip.
Open a terminal in the project directory and activate your virtual environment.
Run the pipeline:
python main.py "https://www.youtube.com/watch?v=dQw4w9WgXcQ" 45 8Arguments
arg1 (required): YouTube URL (wrap in quotes)
arg2 (optional): Clip duration in seconds (default: 45)
arg3 (optional): Number of clips to generate (default: 8)
Examples
# Defaults: 8 clips × 45s
python main.py "https://www.youtube.com/watch?v=abcdef"
# Custom: 6 clips × 30s
python main.py "https://www.youtube.com/watch?v=abcdef" 30 6Important constraint: Total requested time must fit within the source video. Example: For an 8×45s request you need ≥ 360s of source. If the video is only 300s, the run will fail.
After the first run, output/ and temp/ are created automatically.
Format: vertical 9:16 (1080×1920 by default, unless configured otherwise)
Count: as specified by number_of_clips
Location: output/
Q: Do I need an NVIDIA GPU?
A: Not strictly, but a CUDA-capable GPU speeds up processing significantly. CPU-only mode is possible with CPU PyTorch wheels.
Q: The script errors with duration mismatch. Why?
A: Your requested clip_duration × number_of_clips exceeds the video's total length. Reduce duration or the number of clips.
Q: Audio desync or quality issues?
A: Make sure FFmpeg is installed and up to date. Try re-running with fewer concurrent processes if your system is under heavy load.
Q: Face tracking?
A: Not yet. Current version crops from center. Face-aware framing is on the roadmap.
- Face-aware framing / motion tracking
- Smart cut-points via speech/scene detection
- Auto B-roll insertion (royalty-free sources)
- GUI wrapper
- Presets for TikTok/Reels/Shorts export
Contributions are welcome!
Open an issue for bugs/feature requests
Submit a PR with a clear description and testing notes
Follow conventional commit messages if possible
This project is licensed under the MIT License. See LICENSE for details.
v0.1.0 – Initial public release: URL input → 9:16 clips (center-crop), defaults 8 × 45s.