Tiny Desk AI

The clip people are copying

The outstanding tiny desk video, explained

Somewhere around the middle of October 2026, the last few seconds of a Tiny Desk concert started turning up in places that had nothing to do with music: group chats, team channels, and the timelines of people who had just watched their team lose on a Saturday. The clip is fifteen seconds long. The band walks into the chorus of Outstanding, the camera lands on the singer, and the saxophone comes in right at the end.

What changed is who is standing there. The faces are swapped, sometimes one, sometimes the whole band, and the joke lands without a caption because everyone already knows the clip.

This page is about the format itself: what those fifteen seconds contain, why the same section gets reused every time, and what you have to decide before you make your own. If you only want the tool, the button on the home page starts it.

Why it is always the same fifteen seconds

Because the recognition is the whole trick. A remix of a scene nobody remembers needs an explanation. This one does not: the turn toward the camera, the horn coming in, the way the room sounds when the band leans into the chorus. Change the faces and the shot still reads as the thing everyone saw.

It also happens to be a useful length. Fifteen seconds is what TikTok, Reels and Shorts all reward, it fits inside a single send in a group chat, and it is short enough that a face only has to hold for a few beats instead of a whole scene.

The five seats

There are five people in the frame, and the order you send photos in decides which face lands where. This is the part people get wrong, and it costs a whole render when they do.

01

Lead singer

Centre of frame, closest to the camera, mouth open on the first word of the chorus. This is the face people look at, so it is the one your group will recognise first.

Face on, eyes to the camera, neutral or half smiling. A photo where they are already laughing comes out laughing for the whole fifteen seconds.

02

Front keyboard

Left of the singer, turned three quarters away, hands on the keys. They are in focus for most of the clip because the camera starts here.

Head and shoulders, slight angle. Do not send a photo where the person is looking over a shoulder, it reads as a back.

03

Back left

Behind the keyboard, half hidden by the singer when the camera swings. They still show up in the wide shot and in the last beat.

Any clear photo of the face. Distance does not matter here because the model only keeps the face, not the room around it.

04

Back right

Opposite the back left, on the side the saxophone leans into. They get one close pass in the middle of the clip.

Same as above. If two people in your group look alike, keep the order straight, the model does not fix a swapped file.

05

Saxophone

Right of frame, the only person whose hands are visible the whole time. The horn is part of the shot, so their face sits behind it.

Face clear, hands out of the way. A photo holding something comes out holding it through the whole clip.

The numbers are the upload order. Getting the order wrong is the most common mistake, and it costs a whole video.

What you actually get

A vertical MP4, 15 seconds, 720p, 9:16, with the sound already in the file. No watermark, no captions, nothing to add in an editor. One video $4.99, three $9.99, five $14.99. No subscription.

The face you send is the face that comes out, with the caveat that small details drift: glasses, jewellery and a very specific hairstyle rarely survive intact. If a face comes back clearly wrong, that counts as a failed render and the order is refundable.

Before you send a photo of someone else

Use a photo you have permission to use. The usual rule applies: if you would not post the photo on its own, do not put it in a video either. Friends, family, a team, a group of colleagues who are in on the joke, that is the whole use case.

Public figures are a different matter. A video of a real person appearing to say or do something they did not do can get an account banned, and in some places it is a legal problem. The clip is funny because everyone knows it is not real, so label it as AI and keep it that way.

Make one from your own photos

One photo for Solo, five for Five. You pick the mode, send the faces, and the fifteen seconds come back with sound in the file.

Start with a photo
FAQ

Tiny Desk Video Generator FAQ

The practical stuff, answered before you ask.

It is a fifteen second clip from a Tiny Desk concert where the band plays the chorus of Outstanding. People took the last few seconds of it and put other faces into the frame: coaches whose teams just lost, office teams, friends.

Because that is the part everyone recognises. The chorus, the turn to the camera, the saxophone coming in. Change the faces and the joke lands without anyone needing an explanation.

Anyone whose photo you have permission to use. Friends, family, a team, yourself. Public figures are a different matter: a video of a real person saying or doing something they did not do can get you banned or sued, so do not post one that way.

The file arrives with sound already in it. If you want a different track, mute the video in the app you post it from and add the licensed version there. Do not upload the original recording anywhere you do not have the rights to use it.

That is what Five mode is for. Five photos, five seats, in the order you send them. The full band is on screen for the whole fifteen seconds, so the order matters more than it does in Solo.

The free route is a video model plus a prompt, and it usually takes an evening of retries to get a face that holds. Here the shot already exists, so the only thing that can go wrong is the photos you send.

Yes. It is vertical, fifteen seconds and has sound in the file, which is what those three want. Label it as AI-generated. Most platforms have a setting for it now, and it costs you nothing.

A few minutes at normal load. Busy hours take longer. You can close the tab and come back, the render does not depend on the page staying open.