How to Make a Tiny Desk AI Video

Choose a reference session to replace performers in an existing clip, or select Create Your Band to make a new AI performance. Each mode has its own cast options, duration and audio behavior.

This guide covers both paths in Try Tiny Desk AI. Start in the generator, and compare actual generated results and reference templates before choosing a mode.

Prepare your photos for either mode

Use JPG, PNG or WebP files up to 10 MB each and 20 MB combined. Choose clear, well-lit, front-facing photos showing one consenting adult per image. Avoid covered faces, heavy filters and group photos that make the intended person ambiguous. Only upload photos you have permission to use. Photo previews do not automatically assess face visibility or image quality.

Path A: Replace performers in a reference session

1. Choose Soul Session or Coaches Session

Soul Session requests a 10-second video with up to 4 replacement performers. Coaches Session requests a 15-second video with up to 5. Preview their source clips on the examples page. These are fixed reference sessions; this path does not accept an arbitrary source video.

2. Assign photos to the supported cast positions

Place each photo above the performer you want to replace. Soul Session supports the lead singer, saxophonist, keyboardist and female backing vocalist. Coaches Session offers five cast positions across its group, close-up and saxophone shots. Check the mapping: assigning a portrait to the saxophonist requests a different role from assigning it to the lead.

3. Review framing, template audio and credits

Choose 9:16 or 16:9 framing. Soul costs 100 credits; Coaches costs 150 credits. Replacing more supported performers does not add a separate charge. The completed MP4 keeps the selected template's original soundtrack. This mode does not offer custom song uploads or a separate music license.

Path B: Create Your Band from photos

1. Select Create Your Band and a setup

Choose Create Your Band in the studio to generate a new performance. It does not use Soul or Coaches footage as a reference clip. Select a setup with the number of photos it requires:

• Solo: 1 portrait.
• Duo: 2 portraits.
• Band: 3–4 portraits.
• Full Band: 4–5 portraits.

These setup cards select the photo creation mode. Their concept artwork illustrates a possible scene; the generated result and its input photos provide an actual comparison.

2. Assign your cast and review the scene

Upload the required portraits and check each selected role in the studio. Roles depend on your setup. The studio's room reference guides the scenery; it does not supply another cast member. Close-ups can show fewer people than the total band.

3. Choose duration, framing and generated music

Choose 10 or 15 seconds, with 10 seconds selected by default. Request 9:16 or 16:9 framing; this mode defaults to 16:9. A ten-second generation costs 100 credits and a fifteen-second generation costs 150 credits.

Create Your Band includes AI-generated music in its MP4 output. It does not copy Soul or Coaches audio and does not promise “Outstanding” or another commercial recording. Review the generated sound along with the faces and motion.

Confirm, generate and download

For either path, sign in, confirm permission to use the photos, and review the selected mode, duration, cast, framing and credit charge before submitting. Follow the task in My Videos; after a connection error, check its status before submitting again. Download the completed video and review faces, movements, sound and shot timing before sharing. No fixed completion time is guaranteed.

There is no free generation tier. A 10-second video costs 100 credits; a 15-second video costs 150 credits. One-time packs: $4.99 for 100 credits (up to 1 ten-second video); $9.99 for 300 credits (up to 3 ten-second videos); $19.99 for 800 credits (up to 8 ten-second videos). Starter includes one 10-second video. Prices are before checkout tax.

See current credit packs and limits before purchasing. Starter alone funds one ten-second video; a fifteen-second request needs a balance of at least 150 credits. Technical task failures restore charged credits. A completed result can differ from what you expected, and generating another version is a separate paid attempt. Read the refund policy.

Photos and sharing

Photos remain local during setup. Confirming generation sends them to the generation provider and saves the submitted photos and settings in private storage. See the Privacy Policy for retention and deletion requests.

Use photos, reference material and audio you have the necessary rights to use, and label AI-created scenes when sharing. This is an independent tool, not an official NPR product or artist endorsement. Read what the Tiny Desk AI format means or contact support for help.