Skip to main content
Video Replication uses a source video, reference images of a new character, and your instructions to generate a new version. You can optionally provide a voice sample for that character.

Prepare your inputs

Provide HTTPS media URLs that remain accessible during processing. Obtain permission for the source footage, character likeness, and any supplied voice.

Describe the new version

For example: “Replace the presenter with the character in the reference images. Keep the kitchen setting, shot sequence, and pacing. Use the supplied voice sample for the new presenter.” A request generates one version for one character. To make several character variants, send separate requests with distinct idempotency keys. For a product placement in the original scene, use Video Editing.

Use in Navigator

Open Navigator, find this agent, and provide your brief and assets. Continue in the same thread to review progress and results. For AI assistant connections, see MCP and A2A.

Integrate directly

Use POST /v3/originals/video-replication with a Darwin API key carrying tools:execute and an Idempotency-Key. See the API reference for inputs, output, and error codes. Review character consistency, motion, lip synchronization, audio, and rights before distributing the result. The maximum output length is 45 seconds.
Last modified on November 2, 2026