Video Face Swap
Video Face Swap replaces faces in source videos using a reference face image while preserving motion, lighting, and temporal consistency.
Overview
Video Face Swap is designed for video-to-video face replacement workflows. Upload a source video and a reference face image, then generate a swapped video that keeps the source motion and scene context.
Why it looks great
- Video face replacement: Swap a face in a video using a single reference image.
- Temporal consistency: Maintains stable results across frames to reduce flicker.
- Natural blending: Preserves lighting, skin tone, expression, and scene context where possible.
- Simple API workflow: Submit a video URL and a face image URL through the unified OpenOctopus API.
- Production-ready output: Returns a generated video URL suitable for preview, download, or downstream processing.
Limits and Performance
- Input video: Source video URL or uploaded video file.
- Reference image: Face image URL or uploaded image file.
- Target selection: Optional target face index for videos with multiple faces.
- Processing time: Depends on source video length, resolution, and provider queue load.
- Best for: Short clips, social content, creative video editing, previews, and face replacement tests.
Pricing
Pricing
is based on configured video processing billing.
| Billing item | Price |
|---|---|
| Video processing | $0.010 per second |
Billing Rule
Each request is billed by processed video duration according to the configured per-second rate.
How to Use
- Upload or provide a public URL for the source video.
- Upload or provide a public URL for the reference face image.
- Optionally set the target face index when the source video contains multiple faces.
- Submit the request and wait for processing to complete.
- Preview or download the generated swapped video.
Input Parameters
| Parameter | Required | Description |
|---|---|---|
| video | Yes | Source video used for face replacement. |
| face_image | Yes | Reference face image to swap into the video. |
| target_index | No | Target face index when multiple faces are present. 0 usually selects the largest or primary face. |
Output Format
- Generated video URL
- Video-to-video face swap result
- Gateway-hosted asset URL in the normalized OpenOctopus response
- Raw upstream provider response for debugging and traceability
Pro tips for best quality
- Use a clear, front-facing reference face image.
- Use source videos with visible, well-lit faces.
- Keep the reference face angle close to the target face angle where possible.
- Use short clips for faster iteration.
- Set target_index when multiple people appear in the video.
Note
Ensure you have permission to use uploaded videos, images, and facial likenesses. Output quality may vary with motion blur, occlusion, extreme angles, and low-resolution inputs.