Awesome
WORK IN PROGRESS
Update:
Switched the text encoder to use the Comfy native T5 loader as the model is same that's used with SD3 and Flux.
It can be downloaded from the Manager, or from here, to ComfyUI/models/clip:
fp16: https://huggingface.co/mcmonkey/google_t5-v1_1-xxl_encoderonly/blob/main/pytorch_model.safetensors
https://github.com/user-attachments/assets/a9e479dd-83fa-4558-abce-0d533210c66e
Everything should be autodownloaded, but for manual install:
bf16 pruned versions of the transformers models available here:
https://huggingface.co/Kijai/EasyAnimate-bf16/tree/main
Everything else from any of the V3 EasyAnimate repos:
https://huggingface.co/alibaba-pai/EasyAnimateV3-XL-2-InP-768x768
For now the models are using the following structure to allow using the different transformers -models without downloading everything else multiple times.
ComfyUI/models/easyanimate/
├───common
│ │ .gitattributes
│ │ model_index.json
│ │ README.md
│ │
│ ├───image_encoder
│ │ config.json
│ │ preprocessor_config.json
│ │ pytorch_model.bin
│ │
│ ├───scheduler
│ │ scheduler_config.json
│ │
│ └───vae
│ config.json
│ diffusion_pytorch_model.safetensors
│
└───transformer
EasyAnimateV3-XL-2-InP-512x512-bf16.safetensors
EasyAnimateV3-XL-2-InP-768x768-bf16.safetensors
EasyAnimateV3-XL-2-InP-960x960-bf16.safetensors
Original repo: