AI Art Generation Handbook/IPAdapter/ComfyUI

From Wikibooks, open books for an open world
Jump to navigation Jump to search

Note: We are focusing more on IPAdapter for SDXL models here:


GO to Your_Installed_Directory/ComfyUI/custom_nodes/ and on the address bar , type cmd and inside terminal , type git clone https://github.com/cubiq/ComfyUI_IPAdapter_plus


Go to the official huggingface SDXL to download the IP Adaptor Encoder (clip_vision) first at link below :

https://huggingface.co/h94/IP-Adapter/tree/main/sdxl_models/image_encoder


You should find a model clip_vision file named: model.safetensors and download it to directory Your_Installed_Directory/ComfyUI/models/clip_vision and renamed it exactly to CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors


Go back up one level https://huggingface.co/h94/IP-Adapter/tree/main/sdxl_models and download all the models with .safertensors and .bins file type into this directory Your_Installed_Directory/ComfyUI/models/ipadapter (Create folder ipadapter if not present)

It should be:

1) ip-adapter-plus-face_sdxl_vit-h.safetensors

2) ip-adapter-plus_sdxl_vit-h.safetensors

3) ip-adapter_sdxl.safetensors

4) ip-adapter_sdxl_vit-h.safetensors


For the Unified Loader to works, the .bins files (Take note of FaceID not Plus version) are also needed to be downloaded from here

https://huggingface.co/h94/IP-Adapter-FaceID/tree/main

into this directory Your_Installed_Directory/ComfyUI/models/ipadapter

5) ip-adapter-faceid-portrait_sdxl.bin

6) ip-adapter-faceid_sdxl.bin

7) ip-adapter-faceid-plusv2_sdxl.bin

8) ip-adapter-faceid-portrait_sdxl_unnorm.bin


Finally, we need to download the Lora for the FaceID to work

Your_Installed_Directory/ComfyUI/models/loras

9) ip-adapter-faceid_sdxl_lora.safetensors

10) ip-adapter-faceid-plusv2_sdxl_lora.safetensors


References

[edit | edit source]

[1] https://github.com/cubiq/ComfyUI_IPAdapter_plus