Best Photos for AI Rap Videos: What Works, What Fails

With Two photos, each photo is one performer. The left photo is the lead, who raps into the mic; the right photo is the hype partner. Put the stronger performer on the left.

Group photos need cropping. With three friends in the frame, crop each person into a separate photo before you upload, and keep the crop at 300 px or more on each side. If you only have a picture of the two of you together, switch to One photo: both people should stand side by side with their faces clear and about the same size. The person on the left raps and the one on the right hypes.

• Face a window or stand in open shade. Soft light from the front shows the features the model needs.
• Avoid a bright window behind the person. Backlight turns the face into a dark shape.
• Avoid hard midday sun and colored party lights. Strong shadows and tints can carry into the video.
• Hold the camera at eye level, the face toward the lens or turned slightly. A profile hides half the face.
• Keep the eyes open and the face relaxed. The Seedance portrait guide asks for a frontal close-up without expression for its own face references.

Waist-up is the safest choice. The face is large enough to stay sharp and the outfit still shows. A tight head shot leaves the model to invent the clothes and body, so the stage version can wear something you never owned. A full-length shot from far away shows the outfit but can leave the face small and soft; zoom in on the face before you upload, and if it blurs, pick a closer photo. The Seedance portrait guide recommends frontal full-body and close-up images so that faces and clothing stay consistent, which is the same idea.

• Sunglasses, masks and hats that shade the eyes hide what makes a face recognizable. Use a photo without them, or accept that the performer may wear them on stage.
• Beauty filters, face paint and heavy color grades change the face itself. The video starts from the filtered face, not the real one.
• Cartoon, anime or AI-edited versions of a person are a different face. Start from a real photo.
• Photos forwarded through chat apps or screenshotted from video calls are often compressed and soft. Use the original from the camera roll.

• Refused on upload: the message names the problem, such as too small, too narrow or too wide, or not a photo. Fix that one thing and add the photo again.
• Content check: photos and topics are checked before rendering. If a photo does not pass, nothing is charged. Use another photo, without revealing clothing.
• The faces could not be read: use sharper, front-facing photos with one person each.
• Any failed render: the credits go back to your balance on their own. Finished videos are checked before they are shown, and you can close the tab while it renders, which takes about 5 to 15 minutes on Wan 3.0.

Only use photos of people who agreed to be in the video, never a minor without a guardian, and never a public figure in a way that could mislead. Platforms enforce the same line: TikTok’s AI-generated content rules require a label on realistic AI video and do not allow the likeness of anyone under 18, or of an adult private person without permission.

Uploaded photos, songs and clips are deleted 7 days after the render finishes. The finished video stays in My videos until you delete it or your account. The privacy policy adds that photos are resized in your browser before upload, which strips embedded location data, and that photos and videos are not used to train AI models.