For this studio, prepare two separate files with one adult in each. Choose a visible face, even light and an unobstructed upper body.
A short photo checklist
| Look for | Why it helps your brief |
|---|---|
| One person per file | Makes it clear which reference belongs to each side of the duet. |
| A clear, uncovered face | Gives the reference a readable identity without a mask, hand or heavy effect. |
| Visible shoulders or upper body | Shows the outfit and posture you want to carry into a waist-up scene. |
| Enough room around the head | Avoids making a tight crop the only framing reference. |
| A photo you can use | Both people should agree to appear in a synthetic video. |
Both files must be JPG, PNG or WebP, up to 5 MB each. The two upload slots in this preview are required; there is no single group-photo mode.
Different backgrounds are fine for planning the setup. You do not need to paint an orange backdrop into your photos: the scene description supplies that idea.
An original Hotel Lobby style prompt
This example separates identity, scene, framing and movement. It asks for a new orange-booth performance without an existing recording or song. Copy it into a video tool that supports two image references, then adapt the reference labels to that tool.
Use reference image 1 for the adult on the left and reference image 2 for the adult on the right. Create a new, short duet performance in a minimalist orange recording booth. One black microphone hangs between them. Keep their faces, hair and outfits consistent with the references. Use a fixed front-facing camera and waist-up framing. The person on the left makes a small hand gesture while the person on the right responds with a gentle head nod. Keep both faces visible and the background steady. Vertical 9:16, 5 seconds, silent. No text, logos, existing footage or copied choreography.An example to adapt in a video tool.
The studio itself uses a preset, so you do not need to paste this text into our upload form. The underlying Kie Seedance Mini documentation distinguishes reference-based generation from first-frame and first/last-frame workflows; make sure you use the mode intended for your input.
Change one part at a time
Identity: keep it tied to the photos
Say which reference is left and which is right. Avoid adding a different person’s name or a description that conflicts with the photos.
Movement: choose one short exchange
Try a small gesture and a response. Treat busy choreography, crossed arms and fast camera moves as separate experiments rather than putting every idea into the first brief.
Framing: protect both faces
Keep a fixed, front-facing view with both people in frame. Decide whether your brief needs waist-up or full-body framing before asking for detailed movement.
Use the result to choose the next change
| What you notice | One thing to check next |
|---|---|
| A face drifts during the clip | Check whether the reference face is clear. Compare a simpler movement brief. |
| Hands cross or look distorted | Ask for a smaller gesture with space between the people. |
| One person leaves the frame | Make the framing and fixed camera explicit. |
| The moves differ from a source clip | A photo-only prompt does not define that timing. Consider a supported motion-transfer workflow. |
| There is no sound | This studio is configured for silent output. Add an available track in the app where you post. |
Inspect before generating again
Changing the prompt starts a new experiment in your chosen tool. First check the existing task and its charge status if a submission times out or the result is missing.
Read the workflow guide for original generation versus motion transfer, or the FAQ for this preview’s current limits.