Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

T2IAdapter for SDXL #68

Open
ke1ne opened this issue Sep 8, 2023 · 6 comments
Open

T2IAdapter for SDXL #68

ke1ne opened this issue Sep 8, 2023 · 6 comments
Labels
enhancement New feature or request

Comments

@ke1ne
Copy link

ke1ne commented Sep 8, 2023

Maybe, lllyasviel#325

@barepixels
Copy link

barepixels commented Sep 8, 2023

Let hope @lllyasviel or @MoonRide303 add them to Fooocus

From what I read they require less VRAM

@barepixels
Copy link

@MoonRide303
Copy link
Owner

Loading T2I-Adapters in diffusers format is supported by backend (ComfyUI) since yesterday (commit 1e6b671).

Another options are converted safetensors versions from lllyasviel (used by his powerful ControlNet extension for SD web UI, recently updated to support those new models).

I see the potential in supporting more models (I used LineArt, Sketch and OpenPose in SD 1.5 days with ControlNet 1.1), as well as in T2I approach (smaller VRAM footprint - though Control-LoRAs rank 128 are not bad, either - it's possible to use 2 of them on a simple 8 GB VRAM GPU). Problem is that packing them all into UI could be too much - UI feels a bit overloaded, already, and Fooocus (even MRE) should remain simple and easy to understand UI. I'd also want to have some space for IP-Adapters (when they'll be supported by backend).

So, to sum it up: I am considering it, but I would need to figure out clean way to do so, first (from both UI and implementation point of view).

@MoonRide303 MoonRide303 added the enhancement New feature or request label Sep 9, 2023
@barepixels
Copy link

barepixels commented Sep 9, 2023

Currently which XL OpenPose does Comfyui support? Just only 1? Kinda tired of the same poses in XL.

The other day I generated over 300 pics of a model standing in a field of sunflowers. I got 1pic with her hand on top of her head. Rest were similar variations with both hands down. Yeah I know I should try to change in the prompt often. I have been collecting OpenPose stick figure images. It's just so much easier to drop a pose in instead of trying to describe it. Well for me anyway

@ikmalsaid
Copy link

Take your time. I'm still using Comfy to try out new stuffs. Hoping for the best of MRE!

@aimarxjg
Copy link

It can be downloaded on demand. if you need any function, please go to the "author download address" and put it into the corresponding folder to use this function ~ it is convenient to reduce the volume of the compressed package.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

5 participants