How to Make an Outstanding Tiny Desk Video with 5 Photos

What you are making

This workflow turns five individual portraits into a new AI band performance using the Outstanding Tiny Desk reference scenes. It is not a slideshow of your photographs. Each portrait guides one performer across the relevant shots; the reference soundtrack is combined with the finished video.

The result is approximately 17.6 seconds in a vertical MP4. Choose 480p for 10 credits or 720p for 20 credits. Both options make the full sequence. Higher resolution changes the output format, not the promise of perfect face matching or lip-sync.

1. Prepare five separate portraits

Have a photo ready for every role: lead vocals, backing left, backing right, keys and vocals, and saxophone. The lead gets the smile close-up. Keys and the backing singers appear together, and the saxophone has its own shot. These recurring appearances are why a face needs to remain recognizable across more than one camera view.

Use one visible person per image. A clear head-to-waist portrait gives the model useful information about hair and clothing as well as the face. Avoid sunglasses, a hand across the face, a phone covering the body, or a very small face in a group photograph. Read the photo preparation guide before choosing your inputs.

All five role slots require a photo. This workflow does not automatically fill empty band seats, and a group photo does not replace five separate uploads. Choose images you have permission to use.

2. Upload and assign the cast

Open the studio, sign in, and place one portrait into each named slot. Check the role labels rather than relying on the order of files on your device. If you change your mind, remove the image from that role and upload its replacement.

The upload accepts JPEG, PNG and WebP images up to 10MB each. Images are resized and compressed before storage. A successful upload is not a guarantee that the person will be reproduced exactly; it only confirms that the input is ready for generation.

3. Choose quality and check the credits

Select 480p or 720p and read the displayed credit cost before making the video. You need enough credits for the chosen quality. A 10-credit pack covers one 480p video, but it does not cover a 20-credit 720p video. See the credit and pricing guide for pack counts and mixed-quality examples.

4. Generate and return to My videos

Start the generation once. Four scenes are generated, joined in the reference order, and combined with the reference soundtrack. Allow about 20–40 minutes depending on the queue; this is an estimate rather than a deadline. You can close the tab and return to My videos.

A queued or rendering task is still in progress. If the status requests review, contact support with the task ID rather than repeatedly trying to create the same video. A task that ends in failure returns the credits it used.

5. Review the MP4 before sharing

Check each scene: the lead close-up, the group performers, the saxophone, and the ending. Look for missing replacements, facial changes, clothing changes and mouth timing. A completed render can still have visual differences. Our quality expectations guide explains these limits.

Download the MP4 from My videos. The file includes the reference soundtrack; a separate silent file is not included. If you want different audio, mute or replace the audio in your editor. Review the finished post and disclose its AI-generated nature when sharing.