How to prepare photos for an AI rap video
Choose a useful portrait or body reference, identify each performer, and prepare a brief you can check against the finished video.
A useful photo shows the details you expect the video to keep. Choose the performance first. A close-up of someone rapping needs different reference information from a full-body remake of a familiar dance.
Match the framing to the performance
For a face-led singing or lip-sync shot, start with a well-lit portrait facing the camera. Keep the eyes, mouth and jaw visible. DreamFace's singing-avatar guide gives the same starting advice for its photo-and-audio workflow.
For a performance where clothing and body movement matter, prepare a wider reference. Show the jacket, silhouette and limbs that identify the person. A face crop provides no reference for shoes or a full-body pose. Keep the original wider photo so you can make a different crop if the chosen tool requires one.
Prepare the crop that matches the workflow's actual instructions. There is no universal photo size or framing that guarantees a good result across models.
Make identity easy to judge
- Use the original image when available. Avoid a screenshot of a screenshot with a small, compressed face.
- Choose even lighting so facial features remain visible. Heavy filters can remove details you want to preserve.
- Keep hands, hair and props away from the mouth when mouth movement is part of the result.
- Retain an important hat or uniform, but choose an angle where it does not hide the face.
- Use your own photo or a subject you have permission to include. Decide which identifying details must survive before you start.
Pick a primary reference rather than a collection of conflicting outfits and hairstyles. More images only help when the workflow explains how it uses them.
For two performers, write down the roles
If the selected workflow accepts separate photos, prepare one clear reference for each person. If it accepts a shared starting image, check that both people are visible and separated. Do not assume every tool accepts both input formats.
Write down who leads, who responds and where each person stands. During review, check the face and clothing together. A correct face attached to the other person's outfit is still the wrong result.
Here is a planning brief for a hypothetical duo:
Alex wears the red jacket and starts on the left. Jo wears the dark shirt and stays on the right. Alex delivers the opening line while Jo nods and listens. Jo takes the next line. Keep both people in frame and preserve their clothing.
This describes a result to test. A written role assignment does not establish that a model can follow it.
Review the difficult moment first
When reviewing a generated video, inspect the turn of the head, the hand crossing the face and the change of singer. Check whether the person remains recognizable, the mouth follows the intended voice, and the non-singing partner behaves as planned. For a pet or illustrated character, test that input separately. A human example does not establish the same behavior for another body shape.
Choose your two portraits in the studio, sign in and create your rap video. See pricing for credit packs and generation costs.