site banner

Friday Fun Thread for May 19, 2023

Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.

3
Jump in the discussion.

No email address required.

If you want very specific and controllable results, by all means use controlnet. You can inherit silhouettes and poses from reference images, such as a humanoid silhouette that you can make into a robot without the issue of pauldrons.

That, or deleting the offending section and inpainting over and over again till it's fixed.

While SD isn't the most advanced model by itself, it's the extensions that let it put up a fight against Midjourney and the like!

Got it working now. It's fascinating how much the open source community can do, yet the heavyweights can just overwhelm it with computational firepower. Google's high-parameter models could produce readable text on demand in photorealistic images before Stable Diffusion even existed.

https://twitter.com/JeffDean/status/1539743557476663300