omnihuman_create_task
Generate a talking or performing video of a person from one image and one audio clip (OmniHuman 2.0).
Returns a task_id; poll omnihuman_get_task, which persists the finished video (the provider URL expires after one hour). Submission is never retried automatically. Requires background execution (ark_job_submit).
| Requirement | Value |
|---|---|
| Registration | Requires BYTEPLUS_OMNIHUMAN_ENABLED=true and BYTEPLUS_CV_API_KEY. |
| Execution | Background required |
| Required permission | omnihuman:create |
This site lists the full configured capability surface, not the tools enabled on your particular server. Use model discovery and job discovery to check your connection. Credentials are server configuration, never tool arguments.
Supported Models
Section titled “Supported Models”These are supported families and built-in model or service IDs for this site revision, not provider account access or a guarantee that every row is enabled on your server. Check model discovery for configured bindings and service availability.
| Model | Model or service ID | Availability | Capabilities |
|---|---|---|---|
| OmniHuman 2.0 | realman_avatar_picture_omni20_cv |
BYTEPLUS_OMNIHUMAN_ENABLED=true and BYTEPLUS_CV_API_KEY; disabled by default. |
Image plus audio to talking-avatar video; asynchronous tasks polled with omnihuman_get_task. |
This tool has no per-call model argument. The identifier is a Vision AI service req_key, configurable with BYTEPLUS_OMNIHUMAN_REQ_KEY, not a ModelArk model ID.
Ask your connected AI client to use omnihuman_create_task for the operation you want.
The client constructs the arguments and handles tool calls. Review its proposed
operation before approving anything billable. For a non-billable introduction,
start with discovery without generation.
Submit through ark_job_submit with tool_name and the complete target arguments, including its inner input object. Poll ark_job_get using the returned Ark job_id, respecting poll_after_ms. Inspect both terminal status and result.is_error; the original output is in result.structured_content. Do not call this tool in the foreground.
Match the input fields below and the full JSON Schema. See the worked examples for this capability and Background Execution for job/result handling.
An Ark job ID is not a provider task ID. Video and 3D tasks use gen_task_get, VOD tasks use vod_get_task, and avatars use omnihuman_get_task. Save accepted provider task IDs and resume polling instead of submitting a second generation.
Inputs
Section titled “Inputs”The MCP arguments object wraps these fields as {"input": {...}}.
| Field | Type | Required | Description and constraints |
|---|---|---|---|
audio |
string | Yes | Driving audio: public HTTPS URL or seed-media:// artifact. wav or mp3, under 15 MB, 4-300 s; video length follows the audio. |
image |
string | Yes | Portrait of the person to animate: public HTTPS URL or seed-media:// artifact (else media_upload and pass the presigned URL). jpeg/png/webp/bmp/tiff/gif/heic/heif, under 30 MB, width/height 0.4-2.5, each side 300-6000 px. |
prompt |
string or null | No | Optional text steering the scene, movements and camera (zh, en, ja, ko, es-MX, id).; Default: null |
watermark |
boolean | No | Whether the provider watermarks the video.; Default: true |
Outputs
Section titled “Outputs”| Field | Type | Required | Description and constraints |
|---|---|---|---|
provider |
“byteplus-vision-cv” | No | Provider surface.; Default: “byteplus-vision-cv” |
recommended_poll_after_ms |
integer | Yes | Suggested delay before polling omnihuman_get_task, in milliseconds. |
request_id |
string or null | No | Provider diagnostic request ID, when returned.; Default: null |
status |
“accepted” | No | Task accepted and queued.; Default: “accepted” |
task_id |
string | Yes | Pass to omnihuman_get_task or omnihuman_cancel_task. |
Nested Types
Section titled “Nested Types”Field types named above refer to these shared definitions. Optional/null fields, enum choices and bounds are retained in the complete schemas below. Schema validation is followed by runtime capability and security checks.
OmniHumanCreateTaskInput
| Field | Type | Required | Description and constraints |
|---|---|---|---|
audio |
string | Yes | Driving audio: public HTTPS URL or seed-media:// artifact. wav or mp3, under 15 MB, 4-300 s; video length follows the audio. |
image |
string | Yes | Portrait of the person to animate: public HTTPS URL or seed-media:// artifact (else media_upload and pass the presigned URL). jpeg/png/webp/bmp/tiff/gif/heic/heif, under 30 MB, width/height 0.4-2.5, each side 300-6000 px. |
prompt |
string or null | No | Optional text steering the scene, movements and camera (zh, en, ja, ko, es-MX, id).; Default: null |
watermark |
boolean | No | Whether the provider watermarks the video.; Default: true |
Complete JSON Schemas
Section titled “Complete JSON Schemas”Input schema, including the MCP wrapper
{ "$defs": { "OmniHumanCreateTaskInput": { "description": "Input for ``omnihuman_create_task``.", "properties": { "audio": { "description": "Driving audio: public HTTPS URL or seed-media:// artifact. wav or mp3, under 15 MB, 4-300 s; video length follows the audio.", "type": "string" }, "image": { "description": "Portrait of the person to animate: public HTTPS URL or seed-media:// artifact (else media_upload and pass the presigned URL). jpeg/png/webp/bmp/tiff/gif/heic/heif, under 30 MB, width/height 0.4-2.5, each side 300-6000 px.", "type": "string" }, "prompt": { "anyOf": [ { "maxLength": 2000, "type": "string" }, { "type": "null" } ], "default": null, "description": "Optional text steering the scene, movements and camera (zh, en, ja, ko, es-MX, id)." }, "watermark": { "default": true, "description": "Whether the provider watermarks the video.", "type": "boolean" } }, "required": [ "image", "audio" ], "type": "object" } }, "additionalProperties": false, "properties": { "input": { "$ref": "#/$defs/OmniHumanCreateTaskInput" } }, "required": [ "input" ], "type": "object"}Output schema, including nested definitions
{ "description": "Accepted OmniHuman 2.0 task.", "properties": { "provider": { "const": "byteplus-vision-cv", "default": "byteplus-vision-cv", "description": "Provider surface.", "title": "Provider", "type": "string" }, "recommended_poll_after_ms": { "description": "Suggested delay before polling omnihuman_get_task, in milliseconds.", "title": "Recommended Poll After Ms", "type": "integer" }, "request_id": { "anyOf": [ { "type": "string" }, { "type": "null" } ], "default": null, "description": "Provider diagnostic request ID, when returned.", "title": "Request Id" }, "status": { "const": "accepted", "default": "accepted", "description": "Task accepted and queued.", "title": "Status", "type": "string" }, "task_id": { "description": "Pass to omnihuman_get_task or omnihuman_cancel_task.", "title": "Task Id", "type": "string" } }, "required": [ "task_id", "recommended_poll_after_ms" ], "title": "OmniHumanCreateTaskOutput", "type": "object"}Troubleshooting
Section titled “Troubleshooting”- Missing tool: check provider credentials and feature flags, then restart the connection.
- Invalid arguments: keep the correct wrapper and follow current model limits from discovery. A valid schema alone does not prove the chosen model supports the combination.
- A job completes with an error: inspect the original tool result, not only the outer job status. Never automatically resubmit potentially billable work.
- Local output writes from jobs require server-configured
OUTPUT_ROOTS; HTTP cannot write to a client’s local disk.
See configuration, background execution, and artifact delivery.