1. Prepare the two inputs
Start with two separate images you have permission to use. Both people should be adults. Full-body photos help show proportions and clothing. The faces should be large enough to inspect, evenly lit and not hidden by accessories. Backgrounds can differ.
- Choose JPG, PNG or WebP, up to 10 MB per file.
- Keep each dimension at or below 4096 pixels.
- Crop extra people out before uploading.
2. Put each person in the intended role
The left upload represents the left subject and the right upload represents the right subject. Use Swap roles to exchange them. This changes the submitted image order. Removing an image lets you replace it without creating a new task.
3. Match the output to its destination
Use portrait for a vertical edit and landscape for a wider composition. Select 480p for a smaller file or 720p for more pixels. Every duo request is 15 seconds. Generation is silent; the local preview adds reference audio afterward and keeps both versions. Scene directions are saved as part of the request, but the local demo does not perform scene replacement.
4. Review, submit and return
Sign in before requesting an estimate. The estimate uploads and validates both files and returns the credit cost from the server. Submitting reserves credits and creates one persistent job. Keep the existing task if a request is delayed. When it is ready, play the result, download the MP4 and check My videos after a refresh.