Installation Guide for ComfyUI UniAnimate Nodes

In this video I will be showing how to install the comfyUI UniAnimate nodes from the github repository and also through the ComfyUI Manager. All the required lines of code and links are shown in the guide below;

This installation is for ComfyUI_windows_portable_nvidia

The new ComfyUI_windows_portable had Torch: 2.3.1+cu121 which needs xFormers: 0.0.27
(If your Torch version is 2.0.1+cul, then change 'xFormers==0.0.27' in the requirements file which you will find in the github repository we are about to download to 'xFormers==0.0.20')

Click on Code and click download zip

Extract the downloaded file and place in your 'ComfyUI\custom_nodes' folder

To install the requirements, navigate to '\custom_nodes\ComfyUI-UniAnimate-W-main' folder, invoke the CLI from the folder and Enter:

Check for errors after the requirements finish installing. If there are errors, try re-running the installation and feel free to ask for help in the comments section. And of course, ChatGPT can be a great assistant.

Double-Click 'run_nvidia_gpu' to run ComfyUi, and confirm if the UniAnimate nodes loaded successfully.

Go to this huggingface repository and download the five models shown in the video:

Place the models in '\custom_nodes\ComfyUI-UniAnimate-W-main\checkpoints' folder.

Double click 'run_nvidia_gpu' to run ComfyUI again.

Good luck!

[Isi-uT]

In case you can't click on the links in the description, here they are:
ComfyUI Installation: comfy.org

[williamlocke]

GOT IT WORKING! :)

I re-downloaded your respiratory and started with a fresh install of Comfy. Couldn't get it working on my main Comfy install for some reason.

My big mistake was not understanding that Comfy has its own embedded python environment. I wasted hours installing things in CMD and in conda virtual environments, but nothing was working because.... Comfy has it's own embedded environment. I am learning ^_^

Thanks so much for your nodes and for your help! :)

I was able go beyond the max 64 frame limit by editing unianimate_inferency py but then the limiting factor quickly becomes VRAM. I bet with a bit more tinkering I could ...

[buike]

"Can I add a webcam node instead of load a video node?

[ParthKakarwar-bj]

Can i use a different checkpoint like dreamshaper SD1.5 in place of v2-1_512-ema-pruned.ckpt, can this help with Low VRAM i have 4 gb vram.

[toptipss]

When running I get this error: use_libuv was requested but PyTorch was built without libuv support. I hope the author can help me figure out how to fix it. Thank you

[williamlocke]

Can you tell me what versions of the following you are running? I'll try to match them on a fresh installation of Comfy.

CUDA Toolkit:
cuDNN Library:
PyTorch:
Python:
xformers:

You said you spent some time juggling dependencies yourself. Those are just the ones I THINK matter. Please include anything else that your experience tells you are important.

I'm on windows 10.

[ParthKakarwar-bj]

when i installed the custom node, IT MESSED UP MY PYTORCH VERSION....
Can you plz help me how to get it working on my existing comfy UI , This is my system now, Total VRAM 4096 MB, total RAM 23903 MB
pytorch version: 2.3.1+cu121
xformers version: 0.0.27

[숙회지]

The pose is loaded perfect. but "use_libuv was requested but PyTorch was build without libuv support" corresponding runtime error has occurred. I use pytorch 2.3.1 for cuda12.1. I'm on windows 10. Is the version of CUDA the problem? I want use this node.

[Rachelcenter]

and the linked url is cut off in your description underneath the video

[williamlocke]

Thanks for this! I've been at it for a couple hours and it's been tricky. I just can't seem to get the following to all work together...

CUDA Toolkit: Version 11.8
cuDNN Library: Version 8.8
PyTorch: Version compatible with CUDA 11.8 (e.g., PyTorch 2.4.0)
NVIDIA Drivers
Python: Version 3.10 or later
xformers If used, compatible with your PyTorch version
ComfyUI

Does that list of dependencies look right to you?

I install Xformers, comfy tells me that the version of xformers I installed isn't compatible with my version of pytorch. I install a new version of pytorch but then Comfy isn't happy.
I uninstall xformers, then Comfy doesn't see torch anymore. etc etc etc.
Round and round I go.


Gonna take a break and then try again with a fresh install of ComfyUI.
Do you recommend installing a new ComfyUI in its own conda env?