A useful AI motion transfer reference video makes the intended performer easy to follow. Start with a clear solo MP4, a visible face, stable framing, and a portrait with a similar viewing angle. The reference supplies movement and scene structure; your photo supplies appearance. Good preparation reduces avoidable problems but does not guarantee identity or background consistency.
If you need the complete upload-to-export process, follow how to make an AI dance video with your own photo. This guide helps you decide which inputs to use before opening the workspace.
Check the file before judging the performance
MoveFrame AI currently accepts a standard MP4 up to 100 MB and JPG, PNG, or WebP portraits up to 10 MB each. A video filename ending in .mp4 does not necessarily mean its contents are a usable MP4. If an upload cannot be read, open it in your editor and export a standard MP4 rather than just renaming the file.
Play the exported file locally. Confirm that it has the intended picture, duration, framing, and audio. Avoid an unnecessary compression step that makes the performer’s face harder to see. If the file is too large, shorten the test or export a smaller file while retaining readable facial detail.
Compare the footage and portrait
| Check | Prefer | Reconsider |
|---|---|---|
| Face visibility | Eyes, mouth, and jaw remain readable | Face hidden by hands, hair, or a microphone |
| Angle | Portrait resembles the performance angle | Frontal portrait with a mostly profile performance |
| Lighting | Clear facial detail in both inputs | Dark face, strong colored shadows, or clipped highlights |
| Motion | A simple movement you can inspect | Fast spins with persistent blur |
| Framing | Performer remains in view | Face repeatedly leaves the frame |
| Cast | One clear performer per generated shot | Overlapping people needing independent identities |
Pause the reference at the beginning, middle, a turn, and an occlusion. If the face is already unreadable in the original, the generated result is harder to assess. Choose a portrait that is clear at its actual uploaded size; sharpening a tiny screenshot does not restore missing facial information.
Can AI motion transfer handle two people in one video?
Different people can appear across separate scenes with different portrait assignments. The current workspace cannot independently target each person inside the same group or duet shot.
For a two-character edit, record one solo performance for each person and combine those clips into an MP4 before upload. After scene analysis, inspect the timeline and assign the appropriate portrait to each relevant segment. Do not assume the automatic boundaries always match your intended cuts; inspect the preview before generating.
Cropping a duet to one person may help only when that person remains isolated throughout the shot. A second person’s hand crossing the face still creates an occlusion. When isolation is impractical, keep the original group scene or replace the source clip.
Prepare the background and composition in advance
Motion transfer is not a text-driven set builder. Prepare the desired stage or studio in the source video. For a Hotel Lobby-inspired studio edit, arrange the background and microphone in your own footage, with the microphone clear of the face where possible.
For a Chuttamalle-inspired dance edit, start with your own solo performance rather than assuming a couple-dance shot can receive two independent portraits. The song reference does not change the tool’s scene controls.
Clothing, body details, and background can change during generation. If preserving a logo, title card, or exact scenery is essential, keep that segment unchanged or handle it in a separate editing step.
Run a small comparison before expanding
Use the same short reference to compare one input change at a time. For example, replace a blurry portrait while keeping the video unchanged. Review moving frames and scene boundaries, not just a still preview. Record which portrait you used so the next comparison is meaningful.
If the reference remains difficult, the motion transfer troubleshooting guide helps distinguish a visual issue from an upload or task failure. When your files are ready, open the AI motion transfer tool.
Frequently asked questions
Is faster dancing always a worse reference?
Speed alone does not decide suitability. What matters is whether the face remains readable through the motion, without prolonged blur or occlusion.
Should I upload the original music video?
Use footage you recorded or can edit. An official player is a viewing reference and does not supply an editable file or permission to reuse it.
Can a better portrait repair every reference problem?
No. A clear portrait does not remove overlapping performers, missing frames, or severe blur in the source. Simplifying the video may be the more useful change.
