GenX.art DokumentationDigital Human Video
Create a talking-person video from a portrait and audio
Portrait and audio workflow
The current Reference to Video workflow is configured for digital-human generation. It uses a portrait image and an audio track to create a synchronized performance. It is not a general video style-transfer tool.
Workflow
Prepare authorized inputs
Use a clear portrait and audio that you have the right and consent to process. Do not impersonate a person or use a voice without appropriate authorization.
Upload the portrait
Follow the active uploader's file guidance. A front-facing subject with a visible face and limited occlusion generally provides a clearer source.
Add the audio
The audio drives the performance. Use clear speech with limited background noise and follow the format and size limits shown in the creator.
Add presentation guidance
Use the prompt for speaking style, delivery, framing, or scene guidance supported by the current model.
Review and submit
Confirm the inputs, aspect ratio, model, and credit estimate. Processing time depends on the input and provider queue.
Consent and review
- Obtain permission from the depicted person and voice owner.
- Do not create deceptive identity, endorsement, or authorization claims.
- Review lip synchronization, facial motion, and audio before publishing.
- Disclose synthetic media where required by platform policy or law.
Troubleshooting
Face motion is unstable: use a clearer, front-facing portrait with less obstruction.
Synchronization is weak: use clean speech audio and avoid overlapping speakers.
Upload rejected: use the file guidance shown by the current image or audio control; limits can change with the model.
Unexpected framing: choose a compatible aspect ratio and describe the desired presentation in the prompt.
Current scope
This workflow currently uses the configured digital-human model. Its inputs and controls may change as model support evolves; use the active creator as the source of truth.