Playing with "Sygil" WebUI docker and Stable Diffusion

[CyklonDX]

What it is?
Its a WebUI front for Stable Diffusion (mainly), where you can easily start generating your images on your servers.

Link:

GitHub - Sygil-Dev/sygil-webui: Stable Diffusion web UI

Stable Diffusion web UI. Contribute to Sygil-Dev/sygil-webui development by creating an account on GitHub.

github.com

Running Stable Diffusion WebUI Using Docker | Sygil WebUI

<!--

sygil-dev.github.io

(Recommend docker version since its nice)

Docker Config
Extra Params: --runtime=nvidia --gpus all

Vars:
NVIDIA_VISIBLE_DEVICES = all
NVIDIA_DRIVER_CAPABILITIES = all

How does the webui looks like? (Note: there's no auth implemented.)


Now if you have enough vram you can generate bigger images (this chewed through 15GB of vram)

(Used A4000, the generation was quite fast - 30ish sec for 100 steps.

You can generate some beautiful images, more specific arguments describing image helps a ton.

I think if code gets around; it could be used to benchmark cards too, as its vram hungry and does lot of operation on vram. (Not many benchmarks today even try to do so)

 

[CyklonDX]

In case some1 wants to play with it for limited time i'm making this available for 1 week ~ it will stop working on 3/16/2023; obviously if some1 pushes too hard i'll turn it off too.

http://135.148.26.231:8501/?page=stable-diffusion

Regards

 

[CyklonDX]

Has been turned off.

cudo's to some1 who generated quite a few Bruce Willis pictures

(each gen was saved, some of you are quite so with high elf in heaven full body red lips dark sin large boobs ultra realstic...) haven't seen so many elvish milkers.

 

[amalurk]

The face looks not quite like how my brain thinks he actually looks. Like it took his face and forgot some of the detail that makes his face his and so it is more generic. Or no you think his face is accurate?

 

[CyklonDX]

in some pictures it seems more accurate, but in others he looks 2-3 different actors conjoined.

 

[opensourcefan]

I've been using ComfyUI for the past few weeks. The node workflow just works for me. Tried Auto111, it worked but seemed limited.

Stable Diffusion is really cool. Seems to be a decent way to dip your toes into AI Models, processing, iterations etc.