Oh fuck, I want that. Gimme gimme.

The Chlorophyll Queen, hero pose, perfect skin, dramatic lighting, colorful rainbow spectrum cinematography, grand epic, fantastical vista, (Movie Still) (Film Still) (Cinematic) (Cinematic Shot) (Cinematic Lighting) <lora:JuggerCineXL2:1> <lora:xl_more_art-full_v1:1>

And hundreds of thousands of years of evolution pre-training the base model that their experience was layered on top of.

Feeling very Clardic Fug today.

Any reasons why you can’t recommend it?

Interesting, because I saw a 20 point increase between vanilla Firefox and Mercury when testing last night.

That’s not a bad idea. Surely it could be automated within the image. If my ADHD allows me I might take a look at it later. :D

Makeup Artist

Prompt: anthropomorphized bee working as a (makeup artist:1.2), illustrStyle <lora:illustrStyle:1>
Negative Prompt: ac_neg1 ac_neg2 negativeXL_D unaestheticXL_Sky3.1

Looking at the installation instructions, it requires you to run database migrations manually with every image docker image update. Does this mean that running watchtower is going to bork this thing?

I’m not entirely sure how this would help with that, though. Is the model watermarked with your info or something? If you post the result anonymously, I don’t know how you’d track it back to someone that submitted their info here.

Proxmox on physical servers hosting a variety of vanilla Debian installations. I have a physical router running pfsense as well as two HP miniservers running OpenMediaVault.

And another…

I really like this prompt!

That’s never stopped us before.

The problems I’ve had with my RPis have all revolved around the fragility of their SD storage. I got burned one too many times trying to host something important in my house with these things, just for them to get corrupted and lose everything. Backing up these systems was its own nightmare, which failed as much as it succeeded.

What’s wrong with that?

Man, research is advancing at a tremendous pace. So excited about where this is all heading!

I wonder if SD 1.5 would be better suited to this Where’s Waldo style, since generating scenes over 512x512 causes it to start repeating itself, with elements flowing from one 512 region to another. If someone with a ton of VRAM wants to test that theory for me, I’d appreciate it!

I wonder if this model would see an increase in quality if it had the same quantity of data as something like stable diffusion. It is impressive, but I’ve found that it lacks knowledge of some things that stable diffusion understands.

I don’t think you’ve properly thought through the consequences of not considering IP rights for projects with a significant number of contributors. There are absolutely situations in which having a single IP holder is advantageous to having multiple IP holders. Large open source projects might find governance hard when they’re hamstrung by getting consensus from hundreds or thousands of contributors.

And yes, I did read the title and the post. I understood it.