Prepare two separate portraits
Use one adult per image, with an unobstructed face and even lighting. Avoid group shots, tiny faces, strong beauty filters and extreme angles. Choose photos you have permission to use. This build checks image format, file size and dimensions locally; it does not verify identity or consent.
Compose the orange-room scene
Place person one on the left and person two on the right, standing on either side of a microphone. Keep the camera locked, both people fully visible and the background simple. In the planned preview step, check that faces stay distinct and neither person is cropped. Do not proceed with a still you are unhappy with.
Add restrained, alternating movement
Use the approved still as the video reference. Ask one person to gesture while the other nods, then switch roles. Small movements and a fixed camera reduce the number of things the model must keep consistent. This does not guarantee identity preservation or precise lip sync.
Review the clip, then add music
Inspect faces, fingers, clothing and the beginning and end of the clip before publishing. Add a soundtrack you are permitted to use through your editing app or platform. Do not present an AI performance as a real recording of someone without their agreement.
Know the current limits
The creator currently supports photo selection and prompt preparation. AI generation, private file storage and checkout are coming soon. The prompts describe the planned workflow and are not a quality guarantee.