All Changed Merged to main Kijai nodes - Do Not Use

Do not use these nodes any more. Use the nodes at https://github.com/kijai/ComfyUI-LivePortraitKJ. They are more up to date and have all of the MPS support code now.

---

Everything below is kept as an archive

THIS IS FOR M1/M2/M3 Macbooks ONLY. This will not work on windows or linux computers.

The original repo which works with Nvidia GPUs is here: https://github.com/kijai/ComfyUI-LivePortraitKJ

MPS Notes from Grant

If you wish to incorporate these changes into your repo, feel free to open an issue and ask. The commits should be pretty clear, and I also label almost all changes with #HACK so a full text search will work too.

Please let me know if you decide to incorporate any of these changes into your LivePortrait implementation so I can direct people to you repository. I do not intend to maintain this repo.

Some operations are simply not supported on MPS and I didn't rewrite them. Most of my changes are just to .cuda calls and that sort of thing. Many operations are still done on CPU, so don't expect awesome performance.

On my M1 macbook 32GB it takes around 30 seconds flat + 1 second per video frame. 32 frames is 57 seconds, 600 frames is 650 seconds. VRAM usage is not dependent on video length. I never started using swap, but make sure to open up activity monitor and check if you are on a lower-end model. I'm pretty sure it only took 9 GB VRAM but I didn't profile it carefully.

You may also need to run comfyui with PYTORCH_ENABLE_MPS_FALLBACK=1 python main.py (you can also set the fallback as a env variable in other ways), where main.py is the comfyui file. It's possible that it is not necessary macbooks after M1.

ComfyUI nodes to use LivePortrait

chrome_GOS5jyhVAX.mp4

I have converted all the pickle files to safetensors: https://huggingface.co/Kijai/LivePortrait_safetensors/tree/main

They go here (and are automatically downloaded if the folder is not present) ComfyUI/models/liveportrait

Insightface is also required. If you have a working compile encironment, installing it can be as easy as:

pip install insightface

or for the portable version, in the ComfyUI_windows_portable -folder:

python_embeded/python.exe -m pip install insightface

If this fails (and it's likely), you can check the Troubleshooting part of the reactor node for alternative:

https://github.com/Gourieff/comfyui-reactor-node

For insightface model, extract this to ComfyUI/models/insightface/buffalo_l:

https://github.com/deepinsight/insightface/releases/download/v0.7/buffalo_l.zip

Please note that insightface license is non-commercial in nature.