Tiny Desk AI: sing a desk concert with your own face
You behind a cluttered desk under warm lamps, a small band squeezed between the bookshelves, and a held note that ends in a smile.
Tiny desk AI is a 10-second vertical clip made from one photo of you: you stand at a crowded wooden desk with a microphone, a small band plays in the gaps between the bookshelves, the camera pulls back to show the whole room and then pushes in for a held note. It copies the look of the desk-concert trend, not any real show.

A clear, front-facing photo. The template keeps your face, hair and skin tone; outfits and the set come from the template.
No nudity, minors or other people's photos without consent — uploads may be checked by automated tools.
About 5 seconds of you in this format.
This video: 80 credits
No subscription needed · Credits never expire · Usually ready in 10-15 minutes · Failed renders are refunded
Desk Concert Solo
Choreography locked · Facial reference enabled
The song, the band and the room are all invented for the clip, so nothing in it comes from a real recording. It is not an official video from any broadcaster, and the mouth movements are the model's own, not matched to a track you upload.
- Input: one clear photo of one adult, face and shoulders visible. No group photos.
- Output: one 10-second vertical 9:16 MP4 with a made-up soul and funk groove, a sung vocal and room sound.
- Price: 80 credits a video; your first small or medium credit pack is 15% off. No subscription is required.
Three clips made with tiny desk AI
Each clip is one run of this exact template from a different AI-generated portrait. They are shown as they came out, only resized, with the faults written next to them.
What happens in the ten seconds
The shot plan is fixed. These are the beats we saw in all three of our clips.
| Seconds | What the camera does | What you do | What to check |
|---|---|---|---|
| 0 to 4 | Medium shot at the desk. | Sing to the room, nodding on the beat. | The face matches the photo. |
| 4 to 7 | Slow pull-back through the bookshelf light. | Keep singing and nod to the band. | Nobody in the band looks like you. |
| 7 to 10 | Push in to your face. | Eyes closed on a held note, then a smile to the lens. | Nothing readable on the desk or the drum. |
The sound comes with the clip: a made-up groove and a sung line. The words are not controlled, so do not count on a particular lyric.
How to put your own track under the clip
The song in the clip is not matched to your mouth, so most people swap in the track they want. It takes three steps on a phone.
- 01
Mute the clip's own sound
Open the MP4 in TikTok, CapCut or any phone editor and set the video volume to zero. The vocal that comes with the clip is a generated one and will fight with a real song.
- 02
Add a track from the app's own sound library
Pick a licensed track inside the app you post from. A song you upload yourself can be muted or removed by the platform, and the original recording behind the trend is a protected one, so we never use it.
- 03
Line up the first beat with the nod
Slide the track so its first strong beat lands near the three-second mark, where the singer nods on the beat, and trim the end at 10 seconds. Mouth shapes will not match the words, so keep the caption short and let the jokes carry it.
The desk concert trend versus our clip
| Detail | The desk concert trend | Our clip |
|---|---|---|
| Who sings | A real artist or a person doing the bit. | You, from one photo. |
| The room | An office desk, shelves and warm lamps. | Our own version of that look, never a copy of a real set. |
| The band | Real players squeezed in behind. | Invented people: a keyboardist, a bassist and a drummer. |
| The song | A specific recording. | A made-up groove with a sung vocal. No real song. |
| The mouth | Matched to the track. | Moves like singing, but not matched to a track you add. |
| The result | A live recording. | A fiction clip with a few invented details. |
Not affiliated with any broadcaster or show. Tiny desk is used here only as the name of the trend.
Which photo to upload for the desk shot
The room is rebuilt around you. In our three clips the face, hair and top came from the photo, while the trousers and the setting were invented.
✓ Use these
One adult, face forward, even light
All three of our portraits were like this, and the face matched through the close-up in every clip.
A top you would wear on stage
A plain shirt or hoodie carried over in all three clips. Busy patterns may drift, though we only tried one stripe.
A hairstyle that is easy to name
Curls, short twists and tousled hair all held for ten seconds in our runs. Long hair in a complicated style is untested.
✕ Avoid these
Group photos
Two faces compete for the singer's place and one may end up in the band.
Sunglasses, hats and hands over the face
The closing shot needs your eyes. A covered face loses the likeness fast.
Children and pets
This template is for adults. We have not rendered a pet on it.
Who to put at the desk
It works best as a gift or a joke about a person, not as a promotion of yourself. These are the uses that suit a ten-second clip.
A parent who has never sung in public
The warm lamps and a held note suit a parent's birthday post. Ask first, and keep it kind.
The friend who refuses to sing
A short clip of them as the lead singer is a classic group-chat joke.
A birthday or an anniversary
Add a one-line caption over the opening beat and a happy-birthday track from the app's library.
Yourself, for a profile video
A ten-second clip makes a fun first post. Treat it as comedy and say it is AI-made.
Questions about tiny desk AI
What is the tiny desk AI trend?
It is a short video format where a person sings behind a crowded office desk while a small band plays between the shelves, lit by warm lamps. It is also made with AI effects: you upload a photo and the clip puts you in that room. Ours is a ten-second vertical version with three beats, a medium shot, a pull-back that shows the band, and a push-in to your face. The room, the band and the song are invented for the clip, and it is a joke format, not a recording of a real performance. It is made to be posted, with your own track added in the editor.
Is tiny desk AI free?
Some visitors see a free watermarked 5-second preview before paying, which covers the first beat at the desk only. It is a way to check that your face reads well at the desk, not a way to see the pull-back to the band or the closing close-up, and it is not offered to everyone. A full 10-second render costs 80 credits, and your first small or medium credit pack is 15% off. Credits do not expire and no subscription is required. When no preview is offered, the price is shown before anything renders.
Can I use the Outstanding song in my clip?
No. The song behind the original desk concert meme is a copyrighted recording, and we never use it or imitate its lyrics. The clip comes with its own made-up groove and a sung vocal that does not copy any real song. If you want a particular track under your post, add one from the sound library of the app you post to, because a song you upload yourself can be muted or removed. Mute the clip's own audio first so the two do not overlap.
Does the clip lip-sync to my own audio?
No. This version does not take an audio file, so the singer's mouth makes singing shapes without being matched to any particular words, and the sung words in the clip are not under your control. For a post, that means a short caption and a track added in the editor work better than expecting a perfect match. If you need real lip-sync to a song you recorded, a different tool is the better fit. For a joke or a gift clip, most people find the generated vocal is funny enough on its own, or they mute it.
Is this an official NPR Tiny Desk video?
No, and we are not affiliated with NPR or any broadcaster or series. Tiny desk AI is the name people use for the trend, which is the only reason it appears on this page. The outputs carry no logo, show name or footage from any real series, and the room and the band are invented. Do not present a clip as a real appearance on a real show, and do not use it to imply that a person was endorsed by one. It is a fiction clip, so say it is AI-made when you post it.
What photo works best for tiny desk AI?
One clear photo of one adult, face forward in even light, with the face and shoulders in frame. In our three runs the face matched through the close-up every time, curls, tousled hair and short twists all held, and a plain top carried over. Expect invented details: new trousers, a ring that was not in the photo, and some blurry text on records or a drum. A full render takes about 10 to 15 minutes in the shared queue, so pick the photo first and check the face in the final close-up.
Ready to try it?
Add a photo in the tool at the top of this page. Nothing is charged until you press a paid button, and the price is on the button.
Back to the tool