Muerrilla

Influencing only the 0th step does not truly achieve what you're after, since the denoiser quickly 'recovers' from that initial bump and keeps doing its best to generate an image...

> @zahand Very nice examples, I'll try your extension when it's available. However, I also like the unique results I get with only initial noise offset, you could easily include...

> Huh? Theoretically, the reason unmodified SD won't generate dark images is that the model assumes _the added white noise_ has zero mean, so the average color of the input...

> No, that "Image Temporarily Unavailable" is what SD 1.5 generates (for many seeds) when you make it denoise an _impossibly_ bright image (by running numbers above 1 through the...

Yo I had actually implemented perlin noise in dreambooth some time ago, but couldn't really interpret what its effect was (besides that of the offset noise) and then I forgot...

Oh shoot! I wrote this yesterday and posted it today. Now I realized it's already been posted. Sorry, duplicate of #8590.

This is a good workaround for saving model trigger words.

Yes! Please! for DDIM too (screw PLMS, right?).

> > Me three. I looked through launch.py and webui.py, but I can't find where it's doing any version checks. > > > Same problem here. But it extends to...

Ok so after going insane searching my computer for a 3.94GB file and not finding any, I realized if I break the code execution mid-download I might get a hint....