[bug]: Ip Adapters for SDXL Starter Model QoL Improvements

[boricuapab]

Is there an existing issue for this problem?

• I have searched the existing issues

Operating system

Windows

GPU vendor

Nvidia (CUDA)

GPU model

No response

GPU VRAM

No response

Version number

v5.10.1

Browser

chrome

Python dependencies

No response

What happened

I'm trying to use the Ip Adapter plus with an sdxl model (juggernaut xl v9) as I saw that you're able to use it with custom xl in the latest invoke ip adapter video.

I installed them from the starter models, but the plus one doesn't work with sdxl, only with sd 1.5

What you expected to happen

Ip Adapter Plus should work with sdxl models.

How to reproduce the problem

No response

Additional context

No response

Discord username

No response

[boricuapab]

The ip adapter plus is greyed out when I use an sdxl model

[boricuapab]

There is also only one sdxl ip adapter model in the starter models section of the UI, do we need to install the ip adapter plus for sdxl a different way now?

[psychedelicious]

That IP Adapter model is for SD-1 models, not SDXL.

[boricuapab]

Ok, then this request would be different, could these be added to the starter models list on the invoke community, I had to spend time going to HF and installing them one by one, but clicking on the + button in the invoke UI is a big time saver

InvokeAI/ip-adapter-plus_sdxl_vit-h
https://huggingface.co/InvokeAI/ip-adapter-plus_sdxl_vit-h/tree/main

InvokeAI/ip-adapter-plus-face_sdxl_vit-h
https://huggingface.co/InvokeAI/ip-adapter-plus-face_sdxl_vit-h/tree/main

Also I think this may be a bug, but when downloading the starter model called
Standard Reference (IP Adapter ViT-H)
This auto downloads the ViT-G clip vision model into this folder path
models/sdxl/clip_vision/IP Adapter SDXL Image Encoder
but this standard reference model doesn't even use that Vit-G clip vision model so it wastes disc space

And the last starter model that should be added to the sdxl starter models list is this one which uses the Vit-G clip vision model

InvokeAI/ip_adapter_sdxl
https://huggingface.co/InvokeAI/ip_adapter_sdxl/tree/main

This last one auto-downloads the clip vision ViT-G model into this other directory
models/any/clip_vision/IP Adapter SDXL Image Encoder

This last ip_adapter_sdxl is hard coded to use the clip vision model that auto-downloads into the any folder

Lastly also when using the ip_adapter_sdxl, could it auto change the clip_vision option on the canvas, all other models use the ViT-H, but this one errors if it's not explicitly changed to ViT-G.

Thanks