#ai experiments

[therobotmonster]

What is a Dog?

Playing around with Dall-E 3. It has some very impressive coherence but I think we're going to see the leapfrogging happen pretty quick.

But to play with that coherence, especially around text, I asked it to make me:

Prompt: a four-panel comic. a cute kid-sized velociraptor dinosaur trying to figure out what a dog is. Cute and quirky.

The results are beautifully surreal. Some examples in varying levels of weirdness:

This is the closest one to land near an actual idea (by chance, it's still a pareidolia engine), as our dinosaur seems to spot a dog, sees that there's both a dog and and anthro dog, and then has an existential crisis.

This cute little guy with the nostril-sized mouth is taken in by an inconsistent dog, but doesn't know what it does.

"is this a dog?" "What things' a dog?" "What's its quality?" AI's comin' for the anticomedy first.

Robot's like "screw the dinosaurs, I want to explore the Goofy/Pluto conundrum"

And this one's a different prompt entirely, but dang I love how it messed up here

(Prompt: a four-panel comic featuring a humanoid velociraptor in 80s clothing dancing, then tripping, and being embarrassed)

Skip, trip... emrrrrassed indeed.

These images are unmodified and were not adjusted with extensive iteration or other methods, as such they do not meet the minimum human creativity threshold, and are in the public domain,

[klongsuriya]

A Lover's discourses by Suriya Bodhiam

AI generated art

[deepdreamnights]

A friendly wizard and style reference.

Midjourney has just released both the version 6 of its niji anime engine and the first version of its "style reference" tool.

Functionally this is a variation of the image prompting system (explained here), in which breaks a submitted image down into the 'token language' the AI uses internally and uses that as a supplement to a text prompt. "Style Reference" (or 'sref') lets you do this with up to three images, only with only the tokens associated with 'style' being drawn upon.

This is not to be confused with style transfer, a much older and very different AI art process.

But what is a style in this context? And how does it affect generation?

Prompt: a blue axolotl-anthro wizard in a red-and-yellow swirl-pattern robe, holding a sheleighleigh made of purple wood and a potion full of glowing green energy drink. A blue-and-green ladybug familiar stands near his feet, white background, fullbody image

Settings: --niji 6, --style raw --s 50 --seed 1762468963

Here, I've tested the same seed and prompt with a number of reference images.

My semiorganized ramblings under the fold

The first thing I note is that style reference affects the gen so much that same-seed/different style ref comparisons are kind of pointless. Way too much of pose, composition and content changes for it to matter, so for future style ref tests, I'm probably going to drop the seeds.

The second thing I note is that there are certain limitations. You need to change up your prompt for things like photography, and the system interprets styles using its own criteria, not ours. If image prompting misinterprets something, so will style ref, but perhaps not in the same way.

This is notable for the one prompted with a scan from the Nuremberg Chronicle (first row). It recognizes that its a woodcut and emulates that general vibe nicely, but MJ is highly tuned for aesthetics, and emulating real world jank and clumsiness is a weak area. This is literally the first printed (european at least) book with illustrations. Every example thereafter is building on that skillset, so the dataset for woodcuts is going to be largely of a higher apparent quality.

In short, with Midjourney, additional prompt work is needed to replicate the look of early jank or intentionally 'ugly' art styles, and even as recent as v6 I've had no luck with things like midcentury Hanna-Barbereesque cheap TV animation styles or shitty 1990s CGI.

Style reference can help, I've gotten some pretty good cheap 80s-90s TV animation looking stuff from v6 niji and style ref in my early tests:

Color observations: Absent specific requests in the prompt, SREF will stick pretty close to the palette and lighting conditions of the referenced image. With such instructions, you get blending, so the one referencing the okapi fakemon (second row from bottom), for instance, has a lot of colors the reference image doesn't have, but they're in similar in vibrancy and saturation.

One limitation, however, is it doesn't apply to the aspects of the gen that come from any image prompts, so it will always blend the style of the style reference with the style aspects inherited from the image prompt, and that is very strong compared to the style ref.

Using the dog as the image prompt, and the TFTM reformatting as the style prompt, and the text prompt: "a cute older yorkie dog sitting on a bedspread", we get the image on the left. Dropping the image prompt weight to .25 gets us the center option, and removing the image prompt entirely produces the one on the right.

I expect this will be patched eventually, or general image prompting may fall out of favor compared to a combination of style ref and the upcoming character reference option, which will be the same thing, but will only reference the tokens associated with the character in the reference image. Depending on how that works that will have a lot of uses.

Stay tuned for more experiments. There's some good potential for freaky, unexplored aesthetics with combinations of multiple style refs and text prompts.

[theothin]

People are talking a lot about fairies these days, aren't they. Or faeries? Is there a difference?

fairies sitting on a rainbow cloud

Pretty straightforward, even if the right image doesn't quite follow the prompt. Also I'm not sure what's going on with those fairy wings sticking out of the cloud-rainbow. Unsurprisingly, these seem to be the tiny kind of fairies.

faeries on a magic boat

Here, I switch to spelling it "faerie", and it seems to go along with them being bigger than the first ones. We're getting a fancy magic environment, but also a lot of distortions in the faeries themselves. I wonder if that's a result of the more fairy-tale style, it'll be interesting to see if that continues.

faerie men and women riding dolphins

Here, I tried to get more gender variation in the faeries. No visible success, but you never know. The anatomy seems less bad than the previous batch, but still pretty wonky in places. It's interesting to see the differences in location and in the color of the dolphins.

Of course, there's a more relevant marine animal to see a faerie interact with. And let's see if we can get a masc faerie while we're at it.

a faerie man feeding a walrus

Here, it seems "man" overrode "faerie" in the AI's logic. Like with previous walrus prompts, the walrus's tusks are pretty messed up, and both images have different issues with the feeding process. Can we do better?

a male faerie riding a walrus over the ocean

These guys might be a bit more faerie-like, but with no wings. The AI seems really resistant to the idea of male faeries - the systems pick up all sorts of associations from the training data, and some of them can be quite troublesome. Let's try one more time to get past this.

winged male faeries leading a walrus through a cave

Now that's a funny little discovery. I've been selecting the Artistic style for these scenes so far, but this time I accidentally selected Anime instead, which seems much more open to male faeries. Also the walruses are much more cartoony and have wings. What does the Artistic style do with this prompt?

winged male faeries leading a walrus through a cave

Huh. Another winged walrus, but this style remains steadfastly opposed to putting faerie wings on men. Also the right walrus looks more like an elephant.

Anyway, we haven't gotten enough out of the faerie/fairy question. Let's revisit that.

fairies lifting a walrus up into the glimmering sky

Back on the fairy spelling, but looking most similar to the ones from the second and third sets, complete with distortions. Also I have to say, I did not expect either of these lifting methods. But let's assume it works out somehow.

fairies and a walrus in a cloud lake

I wasn't entirely sure what a cloud lake would mean, but sure, that's pretty cloudy. It's important to properly care for your walrus, even if you live in the sky.

Finally, let's see the anime style's take on this.

fairies and a walrus in a cloud lake

Are the fairies small, or is the walrus big? The one on the right looks pretty elephant-like again, and its tusks look weirdly floppy. Also it's wearing what looks like a combination of a blanket of clouds, a wig, and what looks a bit like fairy wings. Not sure what's going on there.

I'm also not sure what the fairies wanted with this walrus, but fairies are known for kidnapping humans, so who knows how ethical it is. Maybe they left a changeling walrus behind. Whatever it is, I hope it works out for everyone involved.

[valandhirwriter]

A few Ai Experiments

Most of them are mood pieces, or illustrations for my stories

[slick-devon]

Graphic novel style portrait The AI didn't give me anything nice, so I further mucked it up in Photoshop. Hey, one can experiment can't they?

[autoeroticalgolania]

the greatest lover alive