Two kinds of tool
The meme needs two jobs done: putting new people into the desk-concert scene, and dressing the clip for posting with sound and captions. Editing apps are good at the second job and cannot do the first. The generators below do the first, and all of them hand back a vertical clip that is ready to post.
So the real choice is between a dedicated generator, which is quick and predictable, and doing the AI part yourself with a general model and prompts, which is flexible but takes several paid tries.
Each generator in a sentence or two
Tiny Desk AI (tiny-desk-ai.app) takes one to three photos: the lead singer, the sax player and the keyboard player, with the two backing singers generated. About 15 seconds at 720p, $9.99 per video, no login. A good fit for one to three people.
AI Tiny Desk (aitinydesk.video) takes one to five photos for the lead, keys, two backing singers and sax; seats you leave empty keep the original performers. About 19 seconds, 10 credits from $9.99, with a free still preview before you pay.
Memelry (memelry.com) takes up to five photos for the same five seats and returns a 15-second 720p clip with the original audio for $9.99, without an account. It is a general meme site with a new trend most days.
GoatRap (goatrap.com) is a rap video site with a tiny desk generator for one to five photos at 1,000 credits, about $9.90; band members you do not cast stay as goats, which is the joke.
Tiny Desk MeMe, this site, takes one to five photos for five roles: the lead, two backing singers, the laughing close-up and the sax player. With five photos, the laugh and the reveal land on two more friends; with fewer, the people you add cover the empty roles. It runs 14 seconds on Seedance 2.5 and costs $9.99 for a 480p clip on the Single pack.
Can I make the tiny desk meme in CapCut?
Partly. CapCut templates for the trend give you the “Outstanding” sound, the cut timing and caption styles, but they do not put new people in the band. People who use CapCut either start from an AI clip made elsewhere and finish it there, or post the template as a reaction with themselves in a corner.
Prompting an AI video model yourself
With a general image-to-video model you can describe the cramped office desk, the five performers and the camera, then render. It is flexible but slow: you have to compose a starting frame with all five people, keep the likeness consistent across shots and recreate the laugh and sax timing by trial and error. Expect several paid attempts.
Chat assistants such as ChatGPT can explain the meme, write that prompt or make a still image, but a chat reply will not hand you the finished band video from your five photos.
How to choose in ten seconds
One to three people and you want it today? Pick a tool built for small casts. A full group of five where everybody, including the laughing close-up and the sax player, is somebody you know? Pick a tool that casts five different people. Only adding the sound to a clip you already have? Use CapCut. Want full creative control and do not mind several tries? Prompt a general video model.
