Skip to content

Guide · 2 min read

How to Create a Tiny Desk AI Video in a Few Simple Steps

To make a Tiny Desk AI video, add a clear lead-singer photo, decide which of the four AI bandmates to keep or replace, select your output settings and create one generation task. Review the finished performance before downloading or sharing.

Start creating

By Tiny Desk AI · Last updated: November 1, 2026

01What Is a Tiny Desk AI Video?

It is an AI-generated music performance set in an intimate office-concert-style space. Tiny Desk AI uses image references to create a new scene with instruments and audio. It does not document a real performance by the uploaded people.

02What Do You Need Before Creating?

Use one adult per photo, with a clear face and even light. JPG, PNG and WebP are supported, up to 10 MiB per image. Each side must be 301–5,999 pixels, with width divided by height greater than 0.4 and less than 2.5.

Use your own photos or obtain permission from the people shown. You need an account with verified email for real uploads and generation, and enough credits for a real task.

03Add Your Photos

Add the lead singer first, then use Replace photo on any bandmate you want to swap. Each photo stays with its role, and removing a replacement restores that AI bandmate. Review the local previews before selecting your video settings.

04Build Your Five-Piece Band

  • Lead singer: one photo, required for every video.
  • Backup singer 1 and Backup singer 2: AI bandmates by default; replace either with your own photo.
  • Keyboard player and saxophone player: AI bandmates by default; replace either with your own photo.

05Select Video Settings

Choose 10 or 15 seconds, a 9:16 or 16:9 frame, then select the model and output resolution available for it. Available settings do not guarantee that every image and combination will produce an equally good result.

Start with the same settings when comparing photo choices. Changing the photo, duration, frame shape and resolution together makes it harder to identify what improved or harmed a result.

06Generate Your Video

Review the cast, settings and displayed credit cost, then click Generate video once. Photo upload and checking run automatically before the task starts. Follow its status; if polling pauses, recover the same task rather than assuming it failed.

07Preview and Download

Once successful, play the video and check each face, body, hand and instrument. Review the audio and singing motion too. Download the result using the workspace control; do not rely on a temporary playback address as permanent storage.

08Tips for Better Results

Prefer clear faces, even light and a natural camera angle. Full-body photos provide more direct cues for legs and body proportions than tight portraits. Group scenes need readable references for every participant; adding people is not a guarantee of a better clip.

FAQ

Frequently asked questions

Short, direct answers. The full workflow and photo rules live in the guides.

Does choosing a photo start generation?

No. Choosing a photo only updates the preview. When you click Generate video, upload and photo checks run automatically, followed by generation if those checks succeed.

How long can the video be?

Choose 10 or 15 seconds. These are the current duration choices.

Can I retry when a task appears stuck?

First query or recover the original task. Creating a new request can create an additional paid task; a temporary query failure does not settle the original outcome.

Keep exploring

Desk is set

Bring your cast to the desk.

Add one lead-singer photo, keep or replace the four AI bandmates, and create only when you are ready. Nothing is generated on page load.