Photo guide / keep each person recognizable
How to Make Your Friends Dance Together Using AI
The hardest part of a friends-dance request is keeping the people distinct while they move. Begin with one clear photo per friend, ask their permission, and use visible details to describe the intended arrangement. Then check every person across the finished video.
Separate photos or a single group photo?
Separate photos usually let you choose a clearer face and more useful body framing for each person. A group photo can still be a source, but crop a distinct reference for each intended dancer. Keep only that person where possible. If someone is partly hidden behind another friend or too small in the original, a new photo gives the model more to work with.
Compare the crops side by side before uploading. Each should have a recognizable face at a similar scale and enough torso or body information for movement. A sharp crop for two people and a blurry crop for the third can make that third identity much harder to preserve.
Step-by-step: prepare and request your friends’ dance
Step 01Choose photos your friends agree to use
Ask each person for a clear photo and permission to animate and post it. Choose similar views rather than mixing a close-up selfie, a distant body shot and a profile. For a group photo, make a separate crop for each person; stop if the face becomes too small or blurry.
Step 02Crop three separate references without hiding the body
Keep only one intended person in each image and preserve the torso and legs where possible. Do not cut a hand or face at the image edge. Use images of comparable quality so the model is not asked to reconstruct one person from far less detail than the others.
Step 03Describe the left, center and right arrangement
Upload the references together in the generator. In the notes, identify people through visible traits such as clothing color and request a left, center and right order. Do not rely on file order alone or assume the model will always preserve the requested positions.
Step 04Check every person before downloading or posting
Start with a short request if your access allows it. Pause at the beginning, middle and end to check faces, clothes and body count. If two identities blend, simplify the references and notes for another attempt. Review the finished clip with your friends before sharing it.
Write useful left, center and right notes
Use features visible in the reference image. A concise example is: “Three friends together, red jacket on the left, blue shirt in the center, green sweater on the right; stable wide camera, all people visible.” The dance prompt is already included in the generator, so the notes can concentrate on who should appear and how they should be arranged.
Keep the arrangement simple. Asking for one person to cross in front of another, spin behind the center dancer and swap sides adds more opportunities for identity confusion. File order and position notes are not a guarantee of placement. Inspect the actual clip instead of relying on the request.
A practical identity check before sharing
First play the clip normally to judge whether the group reads naturally. Then pause near the beginning, middle and end. Check that the same number of people remains visible, each outfit belongs to the intended person and each face still resembles the right reference. Look especially at the outer dancers when arms cross or the camera changes.
Ask the people pictured to review a clip before you post it. If a friend is uncomfortable with how they are represented, use a different result or keep it private. An entertaining effect does not remove the need for their agreement.
What to change after a weak result
| Problem | Next attempt to consider |
|---|---|
| Two friends look like the same person | Use more distinct and clearer references. Keep each person separate, with less overlap and similar lighting. |
| One friend changes clothes halfway through | Mention a short, visible outfit description and remove conflicting costume instructions. |
| The person at the edge is cut off | Request a stable wide shot and leave room around the body in the references. Check any later vertical crop. |
| The output feels unstable over time | Try a shorter request if available under your access. Longer clips give drift more time to appear. |
| The task has not finished | Read its status and use the saved-task check before submitting another generation. |
Save first, then crop and add audio
Download the MP4 and keep an original copy. When preparing a vertical post from a wider result, choose a crop that retains all the dancers; a centered crop can remove the people on the sides. If the trio cannot fit clearly, use a wider presentation instead of enlarging faces until bodies disappear.
Add audio available for your intended use and review the full edit. See the TikTok music guide for sound selection and alignment. For access details, email Billycui50@gmail.com or read the trial page.
Frequently asked questions
Can I make three friends dance from one group picture?
Use the picture as a source for three separate references if each friend can be cropped clearly. When a face is tiny or a body is hidden, ask for a separate photo instead.
Will left, center and right stay fixed?
Position notes can express your intended layout, but the model can still move people or swap positions. Check the output rather than treating the requested order as guaranteed.
How can I reduce blended faces?
Use clear separate images, similar face sizes, distinct visible outfits and simple scene notes. A shorter request may be easier to review, but no setup guarantees identity consistency.
Can I post the result without asking my friends?
Get their permission to animate and share their photos. Review the generated representation with them and follow the destination platform’s current AI-content disclosure requirements.