Bytedance · Utility

OpenClaw x OmniHuman

OmniHuman belongs to the Bytedance catalog on RunAPI and exposes audio-to-video generation plus human-identification and subject-detection helpers through unified API, SDK, CLI, Bruno, and docs surfaces.

3 variants from $0.32 / second Commercial OK
Get API Key View OpenClaw Docs

Prerequisite: npx runapi mcp install

Prompt

Prompt models

Use omnihuman-1.5 when it matches the task: Audio-driven talking-head video from a portrait and driving audio.

You have access to RunAPI task tools for OmniHuman.

Available OmniHuman models:
- omnihuman-1.5: Audio-driven talking-head video from a portrait and driving audio
  endpoints: /api/v1/omnihuman/audio_to_video
  request fields: model, source_image_url, source_audio_url, prompt
- omnihuman-1.5-human-identification: Helper task that identifies human regions in a source image
  endpoints: /api/v1/omnihuman/human_identification
  request fields: model, source_image_url
- omnihuman-1.5-subject-detection: Helper task that returns subject masks for OmniHuman generation
  endpoints: /api/v1/omnihuman/subject_detection
  request fields: model, source_image_url

Use the model ID and endpoint that match the user's request.
Example prompt: Generate a 720p talking-head video from this portrait and voiceover, using a subject mask to keep the speaker isolated.
/api/v1/omnihuman/audio_to_video: Submit the task, poll for its status, then verify the completed output. POST /api/v1/omnihuman/audio_to_video and save the returned Task id.
GET /api/v1/omnihuman/audio_to_video/<task-id> until status is completed or failed.
Verify the completed response contains the endpoint-specific output documented in the API reference.
/api/v1/omnihuman/human_identification: Submit the task, poll for its status, then verify the completed output. POST /api/v1/omnihuman/human_identification and save the returned Task id.
GET /api/v1/omnihuman/human_identification/<task-id> until status is completed or failed.
Verify the completed response contains the endpoint-specific output documented in the API reference.
/api/v1/omnihuman/subject_detection: Submit the task, poll for its status, then verify the completed output. POST /api/v1/omnihuman/subject_detection and save the returned Task id.
GET /api/v1/omnihuman/subject_detection/<task-id> until status is completed or failed.
Verify the completed response contains the endpoint-specific output documented in the API reference.

Public Versions and Endpoints

Model ID Endpoints Price Catalog
omnihuman-1.5
/api/v1/omnihuman/audio_to_video
$0.32 / second Model detail
omnihuman-1.5-human-identification
/api/v1/omnihuman/human_identification
Free Model detail
omnihuman-1.5-subject-detection
/api/v1/omnihuman/subject_detection
Free Model detail

Verify

Poll until the task reaches a terminal status

Select <model-id> to generate verification commands.

Configuration

Guide endpoint: <endpoint>

Select <model-id> to generate a request with the endpoint's public input contract.
How it works

Get Started in 3 Steps

  1. Choose a model ID

    Select a public catalog model ID and review its endpoint and current starting price.

  2. Configure RunAPI

    Set RUNAPI_API_KEY before making the endpoint request.

  3. Verify the result

    For asynchronous endpoints, poll the same endpoint until the Task reaches a terminal status.

What to Build with OpenClaw + OmniHuman

  • Product and marketing clips

    Create short promotional clips for product pages, ads, and landing pages from a written brief.

  • Social media shorts

    Produce short clips for social feeds and try several takes of the same idea before choosing one.

  • Image animation and video edits

    Animate still images or change existing footage where the model offers image-to-video or video editing endpoints.

Why Use OmniHuman Through RunAPI + OpenClaw

  • 3 variants, one API key

    Use one RunAPI connection to choose among the live model variants without changing your integration.

  • Clear usage pricing

    See current catalog pricing before you send a request, with no subscription or minimum spend required.

  • Automatic task workflows

    Submit, poll, and collect asynchronous results through a consistent task workflow without writing manual polling code.

OpenClaw + OmniHuman Questions

What inputs does OmniHuman audio-to-video require?

Audio-to-video requires a source image URL and a source audio URL. Optional mask URLs, prompt guidance, resolution, fast mode, seed, and callback URL can refine the task.

What does human identification return?

Human identification analyzes one source image and returns image outputs that identify human regions for review or preprocessing.

What does subject detection return?

Subject detection analyzes one source image and returns mask outputs that can be used as `mask_urls` in an OmniHuman audio-to-video request.

When should I pass `mask_urls`?

Pass mask URLs when you want stronger subject isolation or when a prior subject-detection task produced masks for the same source image.

What output resolutions are supported?

OmniHuman audio-to-video supports 720p and 1080p output; 1080p is the default when the request does not specify a resolution.

Which model ID should I use?

Choose a public model ID from the version table. Each ID exposes the endpoints shown for that version.

Does this guide configure a chat model?

No. This Model Line uses the endpoint workflow shown here and is not presented as an agent chat model.

Start using OmniHuman with OpenClaw

Get API Key Read Quickstart

Building with a team?

We're here to help with enterprise setup, integrations, and technical questions.

Contact Us