PROMPT · Video Generation
Clay Stop-Motion Cat on a Skateboard Prompt
A ChatGPT Image 2.5 prompt that turns an orange-and-white clay cat — pixel sunglasses, studded collar, cigarette in mouth — into a 24-frame stop-motion skateboard loop, delivered as a finished GIF in one shot.
Saw a tweet from Nett0 (@nett0eth) that starts with a small confession: the thing ChatGPT Image 2.5 handed back was not what he expected.

He asked for a strange little clay cat in stop motion. Sunglasses. A cigarette in the corner of its mouth. Riding a skateboard.

That was the whole ask. What came back was the entire GIF, already assembled. Not a storyboard, not a pile of frames to stitch together yourself — the finished loop.
His words: it delivered the whole gif ready, on its own.
Where the prompt lives
The prompt is short. That is the interesting part. He dropped it in the first reply for anyone who wants to replicate it, and it reads less like a wish and more like a shot list.
Prompt:
Create a 24-frame clay stop-motion loop of an orange-and-white cat riding a skateboard. Handmade plasticine texture, black pixel sunglasses, studded collar, cigarette in mouth. The cat rides, crouches, pops, jumps, flips the board, catches it, lands and loops back. Square 1:1, solid off-white background, full character always visible, no cropped ears/tail/skate, no frame leakage, only a subtle horizontal ground shadow. Slightly choppy handcrafted motion, not smooth CGI. Look at what is actually pinned down in there. Frame count: 24. Aspect ratio: 1:1. Background: solid off-white. Shadow: one subtle horizontal line on the ground. And then the negative constraints — no cropped ears, no cropped tail, no cropped skateboard, no frame leakage. The cat stays fully in frame the whole way through.
The action beats are spelled out too. Rides, crouches, pops, jumps, flips the board, catches it, lands, then loops back to the start. That sequence is what makes it read as a loop instead of a clip that just stops.
The line that does the most work
One clause in there does a lot of quiet work: slightly choppy handcrafted motion, not smooth CGI.
Claymation in real life has a stutter to it. Hands moved the puppet a little bit at a time between exposures, so the motion never comes out glassy. Ask a model for smooth motion and you get something that looks like a render. Ask for the slight chop and you get something that looks like someone actually touched it.
Plasticine texture, handmade — those two words stack with it.
Not everyone got the same result
The replies are worth including, because the outcome is not uniform for everyone.
One person (@bobbygenneric) came back with: here it did not deliver the gif, just the sprites. So frames came out, but no assembled animation.
Nett0's answer is the practical bit: he had only asked it to convert sprites into a gif. On the work version it does better, but on his end it worked even inside the chat.
Two things worth pulling out of that exchange. First, rendering the individual frames and assembling them into a playable loop are separate steps, and the model does not always do both. Second, it points at why the prompt is written the way it is — those frame-count and no-leakage constraints exist to keep the output assemblable at all.
Numbers
As of September 9, 2026:
- Main tweet: 528 likes, 37 retweets, 507 bookmarks, 29,621 views
- Prompt reply: 38 likes, 79 bookmarks, 2,462 views
The bookmark count sits close to the like count on both — 507 against 528 on the main tweet. People are saving this one rather than just nodding at it.
Original tweet: https://x.com/nett0eth/status/2097683441298944340
Prompt (author's own reply): https://x.com/nett0eth/status/2097683445409296829