- Python 3.8+
- Linux
- NVIDIA GPU + CUDA CuDNN
python -m venv interpolation_env
source interpolation_env/bin/activate
pip install -r requirements.txt
docker build -t interpolation_env .
Given an input video and input and target timestamps, run the following command for frames interpolation.
python src/main.py \
<start of input timestamp> \
<end of input timestamp> \
<start of target timestamp> \
<end of target timestamp> \
--input_video <path to the video> \
--gpu \
--fp16
An example would be as following:
python src/main.py 16 30 5 10 --input_video assets/HeyGen.mp4 --gpu --fp16
And using docker:
docker run --rm --gpus '"device=0"' -it -v $PWD:/app interpolation_env python3 src/main.py 16 30 5 10 --input_video assets/HeyGen.mp4 --gpu --fp16
The final output video will be saved in assets/output
.