Tried turning another failure into an alien landscape and got a cloud city.
Has anybody else tried the charcoal drawing setting yet? It can produce amazing results.
Most of those links have other published images, too.
I also found that adding “family” to prompts gives interesting results. Robot family picnic Renaissance painting. Robot family portrait. Or just robot portrait.
(Robots and zombies and creepy dolls are other things to use to avoid the AI’s inability to generate acceptable normal humans. Here are two zombie portraits and a zombie family portrait.)
Thanks, I hadn’t tried that one yet.
Here’s a neat one I just made. There was very high variability depending on seed number, so the starting image is from a private mode batch.
https://creator.nightcafe.studio/creation/ixTuC0MlTIhfGagW1WiY
After seeing some reasonably successful videos, I decided to give it another try.
I wish you could see the generated still photo before deciding if you want to make a video.
I’ve done charcoal drawing a few times when I wanted black and white.
I also, in one of my experiments, tried the charcoal drawing style, with the prompt “green”. The prompt did in fact override the style, and it ended up green (with a texture, oddly enough, that looked like blue jeans).
You could also try adding grayscale/greyscale to the prompt.
Oh, and here’s my charcoal “Green”:
And for comparison, “Green grayscale”:
I tried something yesterday with charcoal and greyscale, and it was still in color. But tge mixed color and greysxale is actually pretty cool.
Played around with Grayscale and Color. Neat effect. Went with spooky themes to play into the black & white part
Eldritch Haunting | Red | Grayscale
I used an interior photo of the Palace of Facets in Moscow as my base for this
Eldritch Portal in a Spooky Church | Cyan | Grayscale
Upscaled once and “God Rays” added as a tag
Pandora’s Box | Yellow | Grayscale
Spooky Swamp | Blue | Grayscale
It took it upon itself to add the green but looks good that it did so.
Do those pipes in the prompt change anything vs. just a stream of text?
They just break up terms to make clear if you want a picture with a turtle and a cop (turtle | cop) or a police officer who happens to be a turtle (turtle cop). I don’t know if they’re always necessary but I just use them as a matter of habit. I’ve also seen people use commas so maybe that works the same but pipes are what I learned first.
This is your fault.
https://creator.nightcafe.studio/creation/GmhUS7DPeonGvxsUqd6v
(BTW, if you clean up the language to sphincters, eyballs, and lips, the images come out differently.)
Looks like it’s eating itself.
Or a hemorrhoid is being reabsorbed.
The AI would be so much better if it could do good human faces. Check out this one I just made. Everything about it is great—except the face.
(It is actually better than most, though.)
Another failure successfully turned into an alien landscape.
That’s a great one.
BTW, that image idea developed from my attempts to paste clip-art of gnomes, elves, and such with photos of mushrooms from my yard. Those never come out the way I want them to, so I tried a gnome alone, calling it both a gnome and a wizard. Even that wasn’t giving me the colorful robe I wanted, but trying for “old man” did—but it was a pretty interesting but abstract profile.
So I just dumped the image altogether and went with a pure text prompt. (That’s not to say that I don’t get some excellent results from crudely patched together source images. More of those in a later post.)
And here are results from doing a gnome group
“Princess | Style of Wallace Wood” failed to emulate Wood’s style at all but giving me a silk gown and accessories abandoned on the palace floor was amusingly (if probably unintentionally) on point
