Make a Tiny Desk AI Performance With Your Photo
Choose a band scene, cast one recognizable lead, and inspect the face and instruments before sharing.

Choose the band look and lead position
In Tiny Desk AI, compare the original band with a family or retro scene. Look for the first clear close-up and the moments when musicians cross in front. Use one approved photo that shows the mouth and shoulders. If the person wears something distinctive, let that be the one styling cue.
Compare the example clips before you generate
Watch the Tiny Desk examples once through, then scrub the preview to see where you would pause. Begin with the clip that frames the feature you want viewers to remember. On a second pass, compare how much space each take leaves around that feature as the subject moves. These times and dimensions describe the preview files; check the generator for current settings.
The Original (540 × 960, 20 seconds)
Use the original setup to find where the room introduces the singer and one instrument together. For your opening, look for the lead and nearby instruments introduced together. Put your own source beside the preview at phone width and check the same lead in the wide shot and close-up. If the main detail is already hard to pick out in a still, compare another source or a preview with more separation before generating.
Family Band (540 × 960, 13.4 seconds)
Watch the group reply and return to the lead. Use those moments to choose where a cutaway belongs in your own edit. To plan the first phrase and the instrument reply, check the frames just before and after the corresponding movement. Note which object or part of the subject first approaches the crop. Keep the last usable frame on either side as a potential trim point.
Seventies Soul (540 × 960, 18.7 seconds)
The period look makes the wardrobe a useful identity cue. Check whether the lead is still easy to recognize when the camera changes size. Work backward from a clean export with the lead still recognizable. Find a frame that could serve as the cover or the start of the next shot. Check it at the intended viewing size. If the important detail disappears there, test the crop in the edit first.
Run the performance from the preset
Pick a take, upload the photo and review the available settings and quote. Keep the example’s performance direction on the first pass. The current editor accepts one starting image. To feature an actual group, create individual takes or add short real reactions from the other people who have agreed.
Prepare the first photo around the action
Put the photo beside the preview and look at what each frame asks the model to show. A face-led opening can work from a close portrait. If the scene then shows footwork, a large arm movement or an important object beside the body, start with the uncropped original or take one wider photo at a similar camera height. Save that file separately so you can go back to the same input after a failed test.
Use a portrait where the performer looks as though they could start a phrase without turning away from the camera. Save one wide still and one close-up for comparison. If the mouth or instrument hands are the weak part of the result, test a shorter reply cut before rebuilding the whole performance.
For this scene, make the same lead in the wide shot and close-up your first check. Open the file at the same width as your draft and see which part of that detail remains clear. If the sample and your source make that check difficult, try a steadier lead close-up and a shorter instrument cutaway. Choose material you have permission to use for the finished audience. With two people in an edit, approve the sources separately and identify which shot introduces each person.
Create a first Tiny Desk performance
- Choose a lead-centered or wider band scene and watch it through. Mark when the lead opens their mouth and whether the instrument players move in front of them.
- Use an approved portrait with the mouth and shoulders visible. Upload it, keep the sample prompt for the first run, and review the available settings and current quote.
- Wait for the result and review first on mute. Follow the lead across the wide view and close-up, then pause at the first instrument shot to inspect fingers and position.
- Listen for the start of the phrase and sound across the cuts. Save a clean export if you plan to add a separate track or recorded group reaction later.
Two short prompts to try with the selected scene
Start with the direction supplied by The Original or the example you picked. For a revision, write what should be visible in shot order. Preserve one subject feature and one main action, then stop. Keep the same scene and source on a prompt test so the result tells you whether the new direction helped.
First pass
Show the lead and one nearby instrument in a small room. Give the singer the first phrase and one musician a short reply. Return to the same lead.
Read the brief once against the preview and decide which detail does the most work. The first look should tell the viewer who is there, and the move should read at normal speed. If your prompt introduces a second setting, a different person or an extra camera move, take that instruction out for this first pass.
One focused revision
Keep the lead steady in the close view. Give the instrument its own short shot with the player's hands visible.
When mouth timing or instrument hands become distracting, test a steadier lead close-up and a shorter instrument cutaway and keep the rest of the setup steady. Compare the same action in both takes, using stills if movement makes the difference hard to see. A working first version is also a reference for your next draft. Save its source, chosen scene and short note together.
Watch the result twice
On mute, track the lead from the widest shot to the close-up. Check the outline of the mouth and the fingers during an instrument shot. Then listen for the timing of the first line and any obvious audio changes. Keep one clean export if you plan to add a platform track later.
For a clean test, use a portrait with an open collar and choose a lead-centered take. Watch on mute to see whether the singer stays identifiable when the band appears. Then listen to the export before assuming a recognizable track or voice carried over.
What to check at the opening, action and ending
Before the action
Use the lead and nearby instruments introduced together to check the same lead in the wide shot and close-up. Compare the generated frame with your source at the same viewing size. Check that the feature you care about is visible before the subject turns, the camera moves or a prop enters. If it is already missing, work on the input or framing first. A later adjustment has more to build on when the opening already looks right.
At the main beat
Around the first phrase and the instrument reply, scrub a few frames in both directions. If mouth timing or instrument hands become distracting, locate the first bad frame and check whether something crosses the subject there. A problem that starts when the hand crosses the face calls for a different correction from a photo that already lacks the right profile. Test a steadier lead close-up and a shorter instrument cutaway. Hold the other inputs steady and repeat the same pause on the new output.
After the move
Find a clean export with the lead still recognizable and decide whether it is a usable final frame or a cut point. Play the full clip at normal speed with the sound muted, then turn the sound back on and play it again. The first pass tells you whether the picture explains the action. The second pass lets you hear whether music or dialogue arrives early enough to make sense without masking the reply. If an action only looks awkward when stepped through frame by frame, judge it in the actual export before rebuilding the shot.
Check the quote and prepare the finished cut
Open the Tiny Desk editor, choose the actual scene and settings, and read the quote shown before submitting. You can check Pavo AI Video pricing for plan details. After submission, use the task state to find the result. Review a completed take before starting another one. If you do need another attempt, write down the one change you are testing.
Keep the clean export and make a separate copy for sharing. Check the cover frame and watch the final copy at normal speed. Add a soundtrack or a caption in the edit where you have the rights to use it. Keep any disclosure appropriate to your destination and audience with the version you send, especially when a recognizable person appears in a fictional performance.
Give the exported file one last check for the same lead in the wide shot and close-up. If the shot needs another pass, keep the exported baseline so the next change is easy to judge. For a separate worked example or a more focused workflow, read Fix a Tiny Desk AI Lead Singer Who Changes Face.
Common questions
What should I change first if the lead looks unfamiliar?
Check the earliest clear face. If that frame is already off, use a sharper portrait with visible mouth and jaw. If the face changes only during movement, choose a scene with a steadier lead angle.
Can my whole group be in the band?
The example can include multiple players, while the editor accepts one starting image per generation. Make separate takes for approved portraits or add real reactions to feature individual friends.