Replace One or Five People in a Tiny Desk AI Video

The Outstanding Tiny Desk AI template has five selectable performer positions: Lead, Sax, Keys, Back left, and Back right. The useful part of this design is that you control the cast at the position level. You can replace one person, create a two-person joke, or map five authorized references to the full group without writing a separate prompt for every subject.

One person: the fastest test

If you want to see whether a particular portrait works, select one position and upload one photo. For example, choose Lead and use one clear friend or pet reference. The other four performers remain in the fixed reference. This is the quickest way to evaluate framing, identity readability, and the overall humor of the result.

One-person replacement also reduces the number of variables. If the output needs improvement, you can change the one source photo instead of trying to understand five simultaneous references. Start with the Tiny Desk AI creator and look for the 1 of 5 state before you submit.

Two or three people: a controlled group meme

A pair or small group is useful when you want a recognizable relationship: two friends, two pets, or one person and one original character. Select the exact positions for the people you want to add and keep the rest unchanged. Use one separate photo per position, even when the final result is intended to look like a group.

Before submitting, read the labels under the selected references. A correct photo in the wrong slot can still create a confusing result. If you make a mistake, remove the photo from that slot and upload the correct reference. Do not rely on the order in which files were selected; rely on the visible role label.

Five people: the full cast

For a full-cast result, prepare five clear photos and assign them one at a time. Use similar levels of clarity across the set. One excellent portrait beside four dark, tiny screenshots makes the final identity mapping less predictable. Similar lighting is helpful, but the more important requirement is that every subject is visible and authorized.

The full-cast workflow is still one fixed performance. Selecting more people changes more identities, but it does not turn the template into an open-ended scene editor. The desk, camera sequence, timing, and instruments remain tied to the reference configuration. Review the Outstanding template details to understand the fixed positions before you create.

What remains unchanged?

The configured workflow is designed to preserve the reference performance structure. Your choices are the selected positions and the authorized photo references. Do not expect the tool to invent a new camera setup, add a sixth performer, write a new performance, or understand a long narrative prompt. A fixed template is valuable because it narrows the creative decision and makes the result easier to compare.

How to avoid a bad group mapping

Use one subject per file, check the position label after every upload, and use the remove control when a file is wrong. Avoid a single group selfie with tiny faces. Avoid asking a photo to represent someone who is not visible. If one person needs a different image, replace only that position and leave the other four untouched.

The current configuration charges one video credit for one to five positions, but check the live checkout before paying. If a generation fails, wait for the original task status and follow the configured credit-refund path instead of submitting duplicates.

For a first experiment, choose one position. For a group-chat meme, choose two or three. For a complete cast, choose all five and review every mapping carefully in the creator.