Digital art creator algorithm website

Not always the style I want, but I’m amazed at the ability to get actual faces with the ‘Dan Witz’ modifier. Nothing else so far gives consistent faces.

The Borg Queen has a crown. :rofl:

BTW, I found that “robot” and “droid” give different responses. Robots are more blocky and droids are more R2-D2ey. (“Cyborg” and “android” get different results, too.)

A Robot crown.

I don’t think I shared this before. This is my favorite borg queen variation. There is a hint of a crown there too.

https://creator.nightcafe.studio/creation/TZFzgwkWvpBv4hz3IzuU

Another of your pet photos?

This borg queen isn’t Trekish, but still interesting.

Dall-E mini borg queen, photorealistic

by Yves Tanguy

cyborg queen, charcoal sketch




pencil sketch



A little Doctor Who-ish maybe.

Definitely looks like the movie one.

These are fanfuckingtastic!

You know what artist I’d like? Thomas Kinkade. Or, if nobody wants to name the pseudo artist, a ‘painter of light’ prompt.

Take a look at that list of artist descriptions I made. Google image search ones that sound like they might produce traits that you want. I tried a few.:

https://creator.nightcafe.studio/u/jBeSWNzpstNXccRiM3xDOPjr1ho2

The evolved Pino Daeni, Dan Witz and Wadim Kashin is my favorite, but I like other ones, too, such as the Pino Daeni, Margaret Keane, Dan Witz and Wadim Kashin or the Pino Daeni, Mark Ryden, Dan Witz and Wadim Kashin. That is barely scratching the surface of potential combinations, though.

(Mark Ryden, Margaret Keane, tronie, and Wanted Poster are all not listed at the site, but I’ve mentioned them all.)

He is on there, both as a listed artist name and as part of the “fantasy” preset, which spells out as “ethereal fantasy hyperdetailed mist Thomas Kinkade”.

This makes me think you might have a misconception of how the AI works, though. With traditional programs, every tool, setting, and menu item is carefully designed for an exact purpose. But these AIs are nothing like that. What they do is create a program that can do pattern recognition. Then they toss several hundred million (not an exaggeration) labeled images from the internet at it. The program then spends a few weeks or months looking at the photos and trying to find definitive details in groups of photos with the same label (planets are usually round, cucumbers are usually green). The AI makes these determinations on its own and nobody, not even the programmers, knows what modifiers have what effects until somebody thinks of it and tries typing it in. That’s why you get weird things like “trending on Artstation” as the name of an effective mod. Nobody named it that, the AI made it and somebody stumbled on it.

So you aren’t limited to any listed style or artist, if you think of something, try it. There may be mods that I’ve tried that nobody else has before. I have no doubt that there are combinations that nobody has tried before. (I’d say that no-one else ever tried the specific combination of Mark Ryden, Kenny Scharf, cloud photography, and drone photography before me and it will never be independently discovered again, for instance.)

Oh, I know. I actually want a ‘paint with light’, um, aesthetic?. Josephine Wall is close . I did not see Kinkade. Let me look again.

This is a Josephine Wall one I did a few days ago.

I just tried “Stone cabin deep in the woods | style of Thomas Kinkade” (coherent renderer)

Meh.

You could improve it with “hyperdetailed colored pencil drawing”, probably.

Thanks, but the credit goes to @Darren_Garrison . I did not modify his prompt.

I can barely run it, let alone train a modified model, so I cannot get it to do anything it does not do on Nightcafe, like understand complex prompts. Well, that is not entirely true; I can mess with all the individual settings as well hack the source code itself, as long as I use the pre-existing models and don’t try anything that takes too much memory.

I upscaled the small one by 8x and the big one by 4x on that Russian supersampler I mentioned. (Trying to do the large one or other already large-ish images by 8x crashes the program with out of memory errors.)

Took the Wall cabin as a starter and ran it with the found Kinkade. Now your talking! Except for the flag. Where’d that come from?

Here is one attempt at the cabin (from noise), using 500 iterations.

there is an artifact looking like a motel sign or something, plus the chimney is certainly messed up (maybe because of my low quality settings)

I love it! Including the ruined chimney. I wonder what makes the AI decide it need to add some Futurama style letters to pictures every now and again.