Skip to main content
The NavTalk Video Synthesis API lets you create digital human videos through a public REST interface. You can submit a task with an avatar, image, or source video, then poll the task status and retrieve the finished video URL when processing completes.

​
What This API Supports

Three REST Endpoints

Submit tasks with POST /api/open/v1/video-compose/submit, then use GET /status and GET /list for polling and task history.

Flexible Audio Input

Provide an uploaded audio file, a public audio URL, or plain text for TTS. Exactly one audio input method is required per task.

Flexible Character Input

Provide an uploaded image or video, a public media URL, or a trained avatarId. Exactly one visual input method is required per task.

Avatar-Based TTS

When using textContent, the API reads the selected avatar’s bound provider and voice automatically. OpenAI, ElevenLabs, and Cartesia are supported.

Asynchronous Workflow

Video generation runs as a background task. The submit endpoint returns a taskId, and you poll for Processing, Published, or Fail.

Public URLs in Responses

The /status and /list responses return full public URLs for source audio, source video, thumbnails, and the final generated result.

​
Authentication

This module uses the same authentication model as the other open APIs.
  • Recommended: pass your API key through the license request header
  • Compatible fallback: pass your API key as the license query parameter

​
Input Rules

Each task must choose exactly one audio source and exactly one visual source.

​
Audio source

  • audioFile
  • audioUrl
  • textContent

​
Visual source

  • characterFile
  • characterUrl
  • avatarId
If you use textContent, avatarId is required. This is because the API needs the avatar’s configured voice and provider to synthesize speech before video generation starts.
  • Use avatarId + textContent when you want the cleanest TTS workflow and voice consistency.
  • Use audioFile + avatarId when you already have recorded speech and want to lip-sync it to a trained avatar.
  • Use characterFile or characterUrl for quick one-off synthesis without training an avatar first.

​
Next Steps