Where I can download the IP-Adapter CLIP G?

It should Clip big Vit G. You have to activate the extension first from what I hear. Should be called something like "sd_forge_ipadapter extension"

@hiben40387 Thanks, I will have a look.

@hiben40387 I selected CLIP-ViT-bigG (IPAdapter) now what should I do?

@norim Select the noob model you downloaded from here in the model section (Put in the controlnet folder in whatever UI you are using). If on reforge sometimes for me it doesn't show up until I refresh a few times. Then load an reference image and you should be good to go.

@hiben40387 Sorry, Reforge extensions dont even work at all. So I cant tried even if I want to.

你可以复制一下我在下面发的图里的节点

@GWH114514 我把模型放进了comfyui_ipadapter_plus/models里，但是好像并没有检测到...

@poisioei 我放在ComfyUI\models\ipadapter里能被检测到

@GWH114514 解决了

exact ! with way more options

comfyui只需要8g，A1111我也爆显存了

same question here.

Put the file you download from here in:

/models/ipadapter

then download:

https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/image_encoder/model.safetensors

rename it to:

CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors

and put it in:

models/clip_vision

I used nodes from:

https://github.com/cubiq/ComfyUI_IPAdapter_plus

and

https://github.com/yolain/ComfyUI-Easy-Use

This is how I use it:

https://files.catbox.moe/7cs5y9.png

@DemiDemi12 Thanks for sharing!

@DemiDemi12 I donwloaded both and it says: I downloaded both and it says:

"ComfyUI execution error: ERROR: controlnet file is invalid and does not contain a valid controlnet model."

When I tried to run it I Came out with something like
KSampler

[Errno 22] Invalid argument

musztarda Error(s) in loading state_dict for ImageProjModel: size mismatch for proj.weight: copying a param with shape torch.Size([3072, 1024]) from checkpoint, the shape in current model is torch.Size([3072, 1280]).我出现了这样的报错

How does this work on a1111? should i just put it in /extensions/sd-webui-controlnet/annotator/downloads/clip_vision/

I downloaded both and it says:

"ComfyUI execution error: ERROR: controlnet file is invalid and does not contain a valid controlnet model."

感谢

did you find solution?

InstantID already works for Noob and Illustrious, search on huggingface shuhe and InstantID 6M. Came out earlier this year, for realistic checkpoints only.

n67113610 That's really cool, thanks for the info!
Did you manage do get it work?
If you managed to, mind sharing what files do i need to download and put in the controlnet folder? Im a noobie at it.

NekoYaSan go to https://huggingface.co/Super-shuhe/InstantID-FaceID-6M/tree/main and download the pytorch-model.bin and put it in models/instantid (I renamed it to pytorch-model-6m) and then download diffusion-pytorch-model from https://huggingface.co/Super-shuhe/InstantID-FaceID-6M/tree/main/controlnet (you can also rename this so you know what it is, renaming either is not necessary. Put this one in models/controlnet

NekoYaSan and yeah it works really well, I suggest setting instantID to begin at 0.5 end at 1. denoise is 0.1, and then the IPadaptor and controlnet strengths are either 0.7 and 0.9 or the other way around.

NekoYaSan they also have a crossfaceID model, which is also instantID and I think it's designed to blend into other model styles rather than be photorealistic. You lose some identity on that one, but by the look of the output I think that was the purpose anyway.

n67113610 Thank you very much bro)

n67113610 It'll be very helpful if you could drop a workflow for it. I am noob at it.

@n67113610 You saved me. Thanks.

@ryuty77292 If you got Instant Id to work, please share a workflow. Also it works for anime faces?

@ryuty77292 Also If you can share some results would be very nice