Workflow exampleConcept frame · not a generated result
Portrait performance
Use a clear face reference with a well-lit source performance for a natural identity transfer.
Try this setupDedicated video tool
AI Video Face Swap
Use a reference face and a source video to create a temporally consistent identity replacement.
Pay per taskMiniMax H3 selected by default
Your source media stays private and is sent only when you submit.
Swap a face in your video
Upload one face reference and one 2–15 second source video. The video model keeps the source performance, camera, clothing, and background while changing the identity.
Only upload media you have permission to use. Do not impersonate, deceive, or replace an identity without consent.
Reference face
Source video
MiniMax H3
80 credits per task
Sign in to upload media and submit this task. Video tools do not include free runs.
Upload all required media before submitting.
Workflow examples
AI video face swap uses a reference face image and a source video. The dedicated workflow sends both inputs to the supported Kie reference-video model and keeps the original performance context.
These visual cards illustrate the workflow with concept frames; they are not generated before-and-after results.
Workflow exampleConcept frame · not a generated result
Short-form creator clip
Keep the original delivery and camera timing while applying a permitted reference identity.
Try this setup
Workflow exampleConcept frame · not a generated result
Character continuity
Maintain the source clip's scene and wardrobe while changing only the visible identity.
Try this setup
Workflow exampleConcept frame · not a generated result
Scene continuity
Keep a moving scene consistent while applying a permitted reference identity across the clip.
Try this setupWhat you can make
- Use one face reference and one source video.
- Keep source motion, timing, lighting, clothing, and background in the edit instructions.
- Review the model, output settings, and rounded credit cost before submission.
How it works
- 1Upload a clear face reference and the source video.
- 2Choose a supported model and output setting for the source clip.
- 3Confirm the visible cost, then create, preview, and download the swapped video.
Frequently asked questions
Do I need to write a prompt?
No. The workflow uses a fixed identity-replacement instruction for this operation; you only provide the reference face, source video, and supported output settings.
What kind of face reference works best?
Use a clear, front-facing image with one visible face and even lighting. The source video should also keep the face visible for consistent tracking.
How long can the source video be?
MiniMax H3 reference videos must be between 2 and 15 seconds. The original upload stays separate and the completed face-swapped video is returned as a new result.