This is a minimal MVP (Minimum Viable Product) for generating short videos by interpolating between two keyframe images using a simple linear blending technique.
Built with Gradio as a frontend and FFmpeg + OpenCV as the backend processing stack.
Visit image2video.paultech to play with demo
- Upload two keyframes (start and end frames)
- Choose number of interpolated frames
- Generate smooth transition video (MP4)
- Gradio-powered UI
- Outputs a
.mp4file in 1920x1920 resolution using FFmpeg
git clone https://github.com/yourusername/keyframe2video.git
cd keyframe2videoUsing venv or conda:
conda create -n keyframe-video python=3.10
conda activate keyframe-videopip install -r requirements.txtIf requirements.txt isn't created yet, you can install manually:
pip install gradio opencv-python ffmpeg-python pillow numpypython app.pyThen open your browser to:
http://127.0.0.1:7860
- A video file is saved to
./temp/<session_id>/output.mp4 - Resolution:
1920x1920 - Format:
.mp4(H.264, yuv422p)
├── app.py # Main Gradio app
├── temp/ # Temporary folder to store frames and output
├── README.md
└── requirements.txt # Python dependencies (optional)
- Replace dummy interpolation with AnimateDiff or FILM for motion-aware generation
- Add ControlNet + Pose conditioning
- Integrate prompt-based animation
- Optional audio/music generation
- Python 3.9 / 3.10
ffmpegmust be installed and added to your system PATH
👉 Download FFmpeg
MIT License. Use freely and modify.
---
Want me to create a `requirements.txt` too? Or generate this `README.md` as a downloadable file for you?