[stavros]

Also, if you don't want to mess around with setting up Jupyter notebooks, I made a service for generating SD images:

https://www.phantasmagoria.me

[asicsp]

See also Stable Horde (https://stablehorde.net/) - crowdsourced distributed cluster of Stable Diffusion workers

[akx]

Automatic1111's fork is pretty good in the no-messing-around sense – on my Windows box (which has an old but decent gaming GPU), all it took was running `launch.bat`, and it set up a venv, downloaded dependencies and wired everything up, and opened the Gradio web UI.

[Multicomp]

Yeah even messing with a venv was too much work for me.

So I used this and it's as simple as extract the download and run the curl | bash script inside it.

Is it safe for enterprise use? Nay.

Is it more convenient than manually futzing with venvs? Yay.

https://github.com/cmdr2/stable-diffusion-ui

[akx]

As my comment says, A1111's fork _sets up the venv for you_, no futzing.

[echelon]

Really cool!

Are you running the GPU instances and model yourself, or are you outsourcing image generation to an API?

[stavros]

Thanks! No, I use banana.dev because I currently have too little traffic to run the GPUs 24/7. That's why the initial generation is currently a little slow, but hopefully that will be improved with some new changes they're implementing soon.