[NVIDIA Only] Advanced Web UI for CogVideo: https://github.com/THUDM/CogVideo
An advanced gradio web ui for generating and editing videos with CogVideo. Supports:
Cogstudio is a single file gradio application cogstudio.py that uses the latest version of CogVideo.
Because it's minimal and blends into the cogvideo repo (instead of existing as a fork), it should take advantage of all the future updates of CogVideo while using the advanced UI.
Enter prompt to generate videos
Upload a video and enter a prompt to transform the video into another video.
Upload a start image and a prompt, and generate a video from the image
Upload a video, select a frame to continue the video from, and generate an extended video.
You can keep extending the generated video to infinity and generate a very long video this way.
Also, it's not just about extending from the end, you can use the slider interface to select precisely which point to continue the generation from.
Once you generate a video, you can send it to:
tabs to add more editing, for example transforming the generated video, or taking the generated video and extending it to make it longer.
[Recommended] Use https://pinokio.computer to install this repo with one click.
First clone the Cogvideo repo:
git clone https://github.com/THUDM/CogVideo
Next copy the cogstudio.py under the CogVideo/inference/gradio_composite_demo
folder.
Go into the CogVideo folder and create a venv:
cd CogVideo/inference/gradio_composite_demo
python -m venv env
Activate the venv.
If you're on Linux:
source env/bin/activate
Or if you're on Windows:
env\Scripts\activate
Install the CogVideo gradio dependencies into the venv
pip install -r requirements.txt
Install PyTorch for CUDA:
pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121
One more thing you need to do is use the latest version of moviepy:
pip install moviepy==2.0.0.dev2
Now start the app:
python cogstudio.py
That's all!