Or a Buttocks with Two Tails.
Oh, I wanted to mention that I had tried this human generator, and after several minutes of not being able to get into the queue at all because it was too busy, and then a four-minute wait once I did get into the queue, I got a person with three legs. Not worth the wait, I think.
My uncle Three-Leg Larry would disagree.
It has a tendency to make hands for feet as well. Bit creepy.
Your aunt is a very lucky woman.
There is a Japanese tradition that on the first night of the new year it is lucky to dream of a hawk, an eggplant, and Mount Fuji. Maybe if you stare at these long enough it will help.
The middle one all the way to the right…
Look at the bottom left one–they both have bedroom eyes.
I didn’t think he was - er - mating with the eggplant - I was thinking that the equipment on this particular eagle looks just like an eggplant emoji.
The bottom left eggplant and eagle are definitely in the process of joining the mile high club.
The handed eagle in the top row and humab legged eagle in the bottom row are also quite noteworthy. (As nightmare fuel)
Well, they are supposed to be from dreams.
Found a test prompt in an article comparing Midjourney 5.2 with 6. You might have fun playing with it.
Draw a marketplace in a futuristic city. To the left of the frame, is a woman with a shopping basket in her left arm. To the right is a street market stall. A robot is behind the stall selling fruit. The robot is purple, and he is holding an apple in his outstretched right arm.
I quickly replaced “draw” with “photo of” and added “grainy disposable camera snapshot with hard flash” to the end. And changing just the word “purple” to a character name is sometimes successful (and sometimes not). So, for example, a modified prompt can be
photo of a marketplace in a futuristic city. To the left of the frame, is a woman with a shopping basket in her left arm. To the right is a street market stall. A robot is behind the stall selling fruit. The robot is Mechagodzilla, and he is holding an apple in his outstretched right arm. grainy disposable camera snapshot with hard flash
I was curious what how Bing/DE3 interprets the concept of Steamboat Willie. I started off with a moment from the Steamboat Willie short:
Photo of Steamboat Willy throwing a potato at a parrot.
No sign there that it recognizes Steamboat Willie being related to Mickey Mouse, but it was interesting output, so I tried:
Photo of Steamboat Willy throwing a parrot at a potato.
Again, no Disney understanding. Next I went less conventional:
Photo of Steamboat Willie steaming a boat willy.
And this time I get Disney related content. So I try the same prompt again and the Disney relationship is literally spelled out (but doesn’t look like the original Steamboat Willie Mickey).
Here’s a third run of the prompt.
I tried that prompt a couple of more times with Disneyish results. I then go back to prompts related to the short film:
Photo of Steamboat Willie standing at a ship captain’s wheel.
Photo of Steamboat Willie feeding hay to a skinny cow.
Photo of Steamboat Willie looking in a goat’s mouth.
Photo of Steamboat Willie playing a cat like a musical instrument.
Photo of Steamboat Willie playing a duck like a musical instrument.
Question for the experts in this thread:
I have a little mini-arcade box that I designed. I’d like to some up with some custom artwork that I’d print on adhesive vinyl and stick to the outside. Something with an 80’s arcade synthwave vibe, but with custom text.
The art would need to fit inside particular shapes, which I can export easily enough. A top marquee, screen bezel, etc. What image generators would work best here? I’m vaguely aware of ControlNet and that it can incorporate a source image, but I don’t know if it can make something that exactly fits in a particular shape. Even better is if I could tell it “fit the artwork in this exact outline, but also use this rough sketch as a general guide”.
Any ideas? I don’t mind paying 20 bucks or whatever for a short-term subscription, if required. Thanks!
Yes, ControlNet could do that.
But you could accomplish the same thing just by using a mask to tell SD where to paint and where not to. I think Night Cafe lets you upload a custom mask (I’ve never needed to do anything that needed more than appropriate hand-drawn masks, so I’ve never tried it.)
Thanks, I finally got the AUTOMATIC1111 UI working with their guide (it’s always Python that causes problems) and the ControlNet extension installed. Looks a bit tricky to use, but I’ll play with it a while to see if it can do what I want. Unfortunately, SD 1.5 looks like it doesn’t quite get what I’m going for, though probably some more prompt engineering would help. It might be tough to convince any of the tools that I just want the flat side of the box, and not a perspective photograph of the whole thing.
I don’t have a local installation of SD (no point, not enough horsepower) but ControlNet does work with SDXL
I only did a couple of tries of ControlNet on Night Cafe, but here’s one:
https://creator.nightcafe.studio/creation/k4OVr5vsVTc4D1smV0JF
One of the "stress tests I try with every AI is zombies doing the YMCA dance. That’s using ControlNet’s pose transfer option using as a template what happened a YMCA set done with stormtroopers I found in a Google image search. (It didn’t do a perfect job with the poses, I probably need a less complex template.)
Here it is with Teletubbies.
https://creator.nightcafe.studio/creation/rIoo6aKpzerXrUTQEoqW
SDXL is a drop-in model for A1111 so you can try that instead.
Additionally, browse the models at Civitai and see if you find something closer to the style you want so the model is doing more of the lifting in your prompt. I saw a couple Synthwave models/Lora that might be promising. Obligatory warning that Civitai is 90% anime, nudes, celebrities, and nude anime celebrities, but there’s other useful stuff in there as well.
Any significant text I’d probably do in a program like Photoshop. None of the AI systems I know of doing a proper job of getting spelling, font, positioning, etc correct.
I can not seem to get Bing/DEe to understand the concept of an axe head with no handle. And some of the “axes” are really weird.
I gave MJ v-6 and Bing a shot at exaggerated perspective. MJ fell short, but Bing pretty much nailed it.
