[Feature Request]: Add preview lora, checkpoint thumbnail

[EasyComfyui]

Is there an existing issue for this?

• I have searched the existing issues and checked the recent builds/commits

What would your feature do ?

Add preview lora, checkpoint thumbnail

Proposed workflow

1. Go to ....
2. Press ....
3. ...

Additional information

[mcmonkey4eva]

the new Model library sidebar tab includes metadata previews. It's a bit of an MVP (only counts metadata in the safetensors header, not those added in eg auto1111 format or anything, and doesn't provide a way to edit). Also doesn't yet account for the model list on a node dropdown

[EasyComfyui]

the new Model library sidebar tab includes metadata previews. It's a bit of an MVP (only counts metadata in the safetensors header, not those added in eg auto1111 format or anything, and doesn't provide a way to edit). Also doesn't yet account for the model list on a node dropdown

i can't see @@

[EasyComfyui]

[mcmonkey4eva]

I think your comfy install is outdated

[EasyComfyui]

Tôi nghĩ rằng cài đặt thoải mái của bạn đã lỗi thời

Too bad it doesn't show up by name

[universorum]

Since ComfyUI released, there in no built-in method to allow user to preview the image of lora/checkpoint. The only way to do that is using a third-party extension. Now we have a built-it sidebar to make selecting model easier but user cannot view the image of the model still. It may be confused when user who have a many models with same concept.

I strongly hope that the UI can show the preview image as A1111 logic. That can let someone like me who often switch the inferencer between A1111 and Comfy more easier to manage the Models.

I noticed that the Comfy API does not return the preview image information in the response. Maybe I should open the issue at main repo?

BTW: Given that this is a very common requirement, I cannot image that this idea never issued at the main repo. But I cannot find the issue by a keyword.

[LukeG89]

@universorum There is the possibility that Model Library will be replaced... Who knows!

@huchenlei proposed to the developer of ComfyUI-Model-Manager and ComfyUI-Image-Browsing to integrate both extensions into core.

• Upstream model manager to ComfyUI core hayden-fr/ComfyUI-Model-Manager#51

[mcmonkey4eva]

Comfy's server currently only returns model metadata, not secondary files, ie if you apply the images in ModelSpec format, Comfy's model browser will see them. If you use SwarmUI to edit model metadata, this is used by default.
Swarm autoreads metadata from other known formats so can compress them for you. When I was working at Comfy Org I had the goal of eventually porting over most of that logic so Comfy could natively read many different metadata formats too, but I didn't get around to that while I was there, so hopefully somebody else will. In the meantime until that's done, you can use Swarm to get all your metadata into standardized format.