Skip to main content

Duet guide

AI Rap Duo Video Generator: Two Photos In, One Duet Out

Create an AI rap duet with a separate photo for each rapper. Pick a shared stage, Make a full 15-second 480p AI rap video for 50 credits.

A duet asks more of an image-to-video model than a solo performance. It must keep two identities recognizable while giving both people room to move.

Our AI rap video generator accepts a separate reference photograph for each performer in two-person mode. The result is one vertical clip, not two independently generated videos stitched together.

We use duo to describe two people sharing the scene and duet to describe the intended exchange between them. Both terms refer to the two-photo workflow here. Two reference photos are required for that mode.

Only upload photos of yourself or people who have agreed to have their images used in this kind of generated performance.

Two rappers, one battle

In a solo shot, the model can devote the frame to one face. In a duet, a close camera move toward one performer may leave the other too small or partly outside the frame.

We generally prefer a scene with enough room for two shoulders and some separation between faces. Side-by-side staging gives the viewer a clearer way to identify who is performing.

Mouth movement is a separate challenge from identity. A prompt can request alternating performance, but it does not guarantee word-perfect lip synchronization or a precise division of lines.

Our current generator does not submit custom lyrics to the provider for synchronization. A moving mouth should not be presented as proof that a person actually performed a particular recording.

Camera movement also affects the exchange. A steady shared shot can preserve the relationship between the performers better than a rapid sequence of close-ups.

We would start with a simple interaction and moderate gestures. If a first result merges hands, swaps features, or loses one face, changing the source images may be more useful than adding a longer prompt.

Pair photographs that agree on light and viewpoint

Choose two clear photographs with similar lighting. A bright outdoor portrait paired with a nearly black indoor image asks the model to invent much more missing detail for one person.

Even illumination across both faces makes a coherent shared scene easier to request. Avoid colored light that hides skin texture or turns one side of a face into a silhouette.

Match the approximate viewpoint as well. Two forward-facing or mildly angled portraits are a more controlled starting point than a frontal photograph paired with a full profile.

We cannot promise a front view reconstructed from a face that is only visible from the side. Glasses, hair, hats, and hands should not cover the features that identify either performer.

Keep framing broadly comparable: two waist-up portraits or two full-body references give the model a more consistent basis for scale. Do not stretch an image to force a match.

Use the original file rather than a screenshot of a screenshot. Our photo preparation notes cover accepted file types and the source-image checks worth doing before any generation.

Plan a short exchange

Watch a public template example to see how the pair is staged. A full 15-second clip leaves room for a shared opening, an exchange, and a closing gesture. Inspect both people throughout the result.

If you will add music in a social app, plan the visual exchange around a short passage you are allowed to use. The generated motion may not land on the words or beats you select.

You can trim an export during editing, but trimming does not create synchronized mouth shapes that were absent in the original generation.

Build a two-person clip in three decisions

Step 1: choose a duet stage. Open the two-photo performance workspace and choose a stage for two performers.

Step 2: upload both references. Use one clear photo per person and confirm that both adults agreed. Do not submit a group photograph in place of the two separate identity references.

Step 3: generate and inspect the export. The full 15-second 480p video costs 50 credits.

When a result is ready, check the entire 9:16 frame for both identities, limbs, and mouth movement. Download only after reviewing it, and add any separate publishing audio through a workflow where you have the required rights.

Duo-specific questions before you submit

Can I use one group photograph instead of two uploads?

Use two separate reference photos for the two-person workflow. A group photograph makes subject selection less clear and does not satisfy the requirement for an individual reference in each upload slot. Start with separate images that identify the intended performers without bystanders.

Will the generator assign an exact line to each person?

No exact line assignment is promised. The duet prompt requests an exchange, but the generator does not receive custom lyrics for word-level synchronization. Treat the output as a generated visual performance and review its actual timing before choosing a final audio passage.

Can the same pair appear in different scene choices?

You can prepare another generation using the same consenting subjects and a compatible scene. Each submission is a new generation with its own credit cost. Identity consistency, positions, clothing, and gestures may vary between results; the references do not create a permanently locked pair of characters.