Odd thing on Night Cafe: images that were 576x384 (or vice-versa) are now 576x448. And 640x384 (ditto) is now 704x384. But they haven’t updated the descriptions on the “create” page.
OK, promptsmiths, anyone want to help me out with this one? I’m trying to create an image for a flyer for my school’s Bike to School Day. My first try at a prompt was “Purple jaguar riding a bicycle”, using NightCafe’s Stable Diffusion. Results were… not great.
Then I thought that maybe the purple was asking too much of it, and it’d do better without the color specified: “A jaguar riding a bicycle”.
OK, that’s two credible jaguars without bikes, and one credible bicyclist without a jaguar, and one only one with both… with the jaguar apparently steering with its chin and pedaling with its forepaws.
OK, third try: “Jaguar riding a bicycle, side view”.
This is not an improvement.
I have over a hundred credits that I can throw at this, if need be, but I’d rather not just flail around in the dark. Any suggestions?
Stable Diffusion 1.5 wasn’t too bad, though the bike is still a bit screwy:
You looking for something cartoonish or “realistic”?
Either one, I suppose, so long as it’s recognizable and verisimilar within its style. Something you’d look at and say at a glance “that’s a jaguar riding a bicycle” is better than something that looks “dramatic”, or whatever.
To be fair, I feel like that’s how a jaguar would do it.
Here are my results from various AIs. Probably unlikely to get something not embarrassing. I think maybe “low poly” images from SDXL on Clipdrop are the most usable.
Personally, though, I’d go with this one:
Plus, you don’t have to use credits. Make one image at a time at Night Cafe. Or make them at Playground.
infill the poop! 
How about something simple:
Only problem with that one is it looks like it’s pooping. 05 lacks that, though, and 24 is also nice in its own way. I think I might just make two versions of the poster. Thanks a ton!
(and since @Jack_Batty posted before I remembered to hit the submit button: That one is also good, but unfortunately the jaguar looks too male-- These particular Jaguars are all girls)
Noooo! The poop is the winning factor!
Dude. Get thee to Playgroundai and play around with infilling. You can paint a mask over any area of an image you don’t like and regenerate it. So it isn’t an all or nothing for an image, you can extensively edit them.
24 is from Kadinsky 2.1.
All my jaguars look like a guy in a furry suit.
Midjourney version 5
Coincidentally I had just yesterday and today been playing around with more tries at “things riding things” and noticed that the AIs kept trying to add bicycles. Including some pretty good ones in Playground v1. So I tried:
Purple jaguar riding bicycle | wide shot | medieval bestiary | Dan Witz, Mark Ryden, Margaret Keane, Pino Daeni, Mab Graves, Carne Griffiths | Tim Biskup, pop surrealism, vinyl toy figurine, Polaroid, Kodachrome, Ektar | Professional photography, bokeh, natural lighting
And got these:
Go to Playground AI, choose Playground v1 as your renderer, and play around with that.
Much better results just posted, but I ran through dozens of iterations at Playground, most of which sucked, but this one popped out and it seems pretty feminine, if not quite well composed:
An in-painted version of the pooping jaguar (via Stable Diffusion local) and three others: one from SD and two from Midjourney with a little inpainting. I quite like #3 but the girlie bike & basket might not befit the students’ athletic prowess. [Edit: Re-reading, I see this is a ‘Bike to School’ event and not a cycling club/team so maybe not as bad]
In Midjourney, “[Animal] riding a horse” often results in a hybrid beast. I got some rockin’ sworses (?) that way. Probably far better than if I tried to ASK for a swan-horse hybrid.
That’s awesome!
The teddy bear will haunt my dreams.