The tiny desk ai prompt
The tiny desk ai prompt is the written instruction an AI video model needs to put a new person into the Outstanding Tiny Desk performance, and it has to solve four problems at once: the face, the room, the timing and the sax solo. On Tiny Desk AI Trend it is already written and tested for every set. Make the video from one photo.
What does the prompt have to solve?
- Keep the person identical to the photo. The big close-up from 3 to 9 seconds fills the frame, so any drift in the face is obvious.
- Rebuild the room. The original is shot in a cramped office corner: shelves of records, books and toys, a red keyboard in front, desk microphones. Without that detail the video becomes a generic stage.
- Follow the beats. Wide shot, laugh close-up, sax solo, wide finale. The four shots have to arrive in order and on time, because the song's sound is laid under them. The shots are listed on the video page.
- Invent the rest of the band. Backup singers and the saxophonist should look like nobody in particular, wear plain clothes and show no logos or readable text.
A written description of this is long, and every sentence in it can break another one. That is why people who find a prompt in a comment section still get something that looks nothing like the trend.
Why do copied prompts fail?
- The prompt is one of several inputs. The motion reference and how the photo is supplied matter as much as the words. Pasting text into a tool that has none of the rest gives you something else.
- The face drifts. Without the right setup the model drifts between the wide shot and the close-up, so by the laugh the person looks like a cousin. Pro and Ultra on Tiny Desk AI Trend hold a face more closely, and the sets are built around that.
- The timing is guessed. The sax solo should land at 9 seconds. A prompt cannot make that happen on its own, so it lands whenever the model decides.
- Every retry costs money. Paid AI video tools charge per attempt, so finding the working wording by trial is the expensive route.
- It changes under you. A prompt that worked last week on one tool version may not on the next.
We do not publish our prompt or settings, on purpose. They are what the price pays for, and a pasted copy would not reproduce the result anyway.
What can you not change?
The room and the sound are fixed for every set: you cannot type a new room or swap the song, because the four shots are timed to the original track. To change the look, pick another set on the sets page. Looking for a CapCut template instead? That is covered on the filter page.
What is built in here?
Each set carries its own tested prompt, motion template and settings, so every render follows the same shot list. The video model is Seedance 2.0 / 2.5 by ByteDance, in tiers from Standard to Ultra; pricing shows what each costs in credits. If a render fails or is blocked, the credits are returned. There is no free tier, and credits never expire.
If you only want the overview, read how to make the video in four steps or the story of the trend. Independent fan-made tool. Not affiliated with, endorsed by or sponsored by NPR, NPR Music, Tiny Desk, Charlie Wilson, The Gap Band, TikTok or CapCut.