> ## Documentation Index > Fetch the complete documentation index at: https://docs.apipod.ai/llms.txt > Use this file to discover all available pages before exploring further. # Create video task (Ark-compatible) > Create an asynchronous video generation task through the Volcano Ark-compatible endpoint. One model name covers text-to-video, image-to-video, and omni reference generation. This endpoint speaks the Volcano Ark (火山方舟) video task dialect. Code written against `ark.cn-beijing.volces.com` works as-is: point the base URL at `https://api.apipod.ai/api/v3`, swap in an APIPod API key, and keep the request body unchanged. Authentication uses the same `Authorization: Bearer ` header as Ark. ## One model name, every scenario Unlike the Ark console, you do not pick separate endpoint IDs per capability. All-in-one model IDs such as `seedance-2.5` and `seedance-2.0-fast` cover every scenario: the task type is inferred from the `content` array — text only is text-to-video, 1–2 frame images are image-to-video, and reference images, videos, or audios are omni reference generation. The three scenarios are mutually exclusive; mixing frame images with reference media is rejected with a 400 error. Un-roled images are frames, so more than two un-roled images are treated as reference images. ## Model names The `model` field accepts all of the following, case-insensitively: | Style | Examples | | -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------- | | All-in-one family IDs | `seedance-2.5`, `seedance-2.0`, `seedance-2.0-fast`, `seedance-2.0-mini` — one model, scenario inferred from `content` | | Volcano native names with a trailing version | `doubao-seedance-2-5-260128`, `doubao-seedance-2-0-fast-260128`, `doubao-seedance-1-5-pro-251215` — normalized to the matching family | | Scenario-suffixed IDs | `seedance-2.5-i2v` — the scenario is pinned by the model itself | ## Parameters Resolution, ratio, duration, seed, generate\_audio, camera\_fixed, watermark, return\_last\_frame, and `tools: [{"type": "web_search"}]` are passed through to the model. Supported values depend on the model; unsupported combinations fail validation exactly like the native `/v1/videos/generations` endpoint. A few Ark parameters are accepted but ignored because the platform does not map them: `service_tier`, `priority`, `execution_expires_after`, `safety_identifier`, `output_format`, `draft`, and `omni_reference_task_type`. The `frames` parameter is rejected with a 400 error — use `duration` instead. ## Callbacks Set `callback_url` to receive a POST when the task reaches a terminal state. The callback body is a task object identical in shape to the query endpoint response, so the same parsing code handles both. See [Webhooks](/webhooks) for delivery and retry behavior of the native callback format used by other endpoints. After creating a task, poll [Query video task (Ark-compatible)](/ark-query-video-task) with the returned `id`. ## OpenAPI ````yaml api-reference/openapi/ark-create-video-task.yaml POST /api/v3/contents/generations/tasks openapi: 3.1.0 info: title: Create Video Task API (Ark-compatible) version: 1.0.0 description: >- Volcano Ark-compatible endpoint for creating an asynchronous video generation task. Point Ark SDKs and clients at this base URL and keep the same request body you would send to ark.cn-beijing.volces.com. One model name covers every scenario: text-to-video, image-to-video (first frame or first and last frame), and omni reference generation are inferred from the content array. servers: - url: https://api.apipod.ai description: Production security: [] paths: /api/v3/contents/generations/tasks: post: tags: - Video tasks summary: Create video generation task operationId: post-ark-create-video-task requestBody: required: true content: application/json: schema: type: object properties: model: type: string description: >- Model name. Accepts APIPod public names (seedance-2.5, seedance-2.0, seedance-2.0-fast, seedance-2.0-mini, and lite variants), Volcano native names with a trailing version (doubao-seedance-2-5-260128), and names that already carry a scenario suffix (seedance-2.5-i2v). example: seedance-2.5 content: type: array description: >- Input content list. Text is the prompt. Images without a role, or with role first_frame/last_frame (max 2), drive image-to-video. Images with role reference_image, videos, and audios drive omni reference generation. Scenarios are mutually exclusive and cannot be mixed. items: type: object properties: type: type: string enum: - text - image_url - video_url - audio_url description: Content item type. text: type: string description: Text prompt, for type text. image_url: type: object properties: url: type: string description: Image URL or data URL. description: Image object, for type image_url. video_url: type: object properties: url: type: string description: Reference video URL. description: Video object, for type video_url. audio_url: type: object properties: url: type: string description: Reference audio URL. description: Audio object, for type audio_url. role: type: string enum: - first_frame - last_frame - reference_image - reference_video - reference_audio description: >- Role of the media item. Un-roled images are treated as first/last frames. example: - type: text text: A cat yawning at the camera - type: image_url image_url: url: https://example.com/first.png role: first_frame resolution: type: string enum: - 480p - 720p - 1080p - 4k description: Video resolution. Supported values depend on the model. ratio: type: string enum: - '16:9' - '4:3' - '1:1' - '3:4' - '9:16' - '21:9' - adaptive description: Aspect ratio of the output video. duration: type: integer description: >- Video duration in seconds. Use -1 for smart selection where the model supports it. seed: type: integer description: Random seed. Supported by a subset of models. generate_audio: type: boolean description: >- Whether to generate synchronized audio. Supported by a subset of models. camera_fixed: type: boolean description: Whether to fix the camera. Supported by a subset of models. watermark: type: boolean description: >- Whether to render an AI-generated watermark. Supported by a subset of models. return_last_frame: type: boolean description: Whether to also return the last frame as an image. callback_url: type: string description: >- Webhook URL. On terminal states APIPod POSTs a task object identical in shape to the query endpoint response. tools: type: array description: >- Tool configuration. Only web_search is recognized, for models that support it. items: type: object properties: type: type: string enum: - web_search description: Tool type. required: - model - content example: model: seedance-2.5 content: - type: text text: A cat yawning at the camera resolution: 720p ratio: '16:9' duration: 5 responses: '200': description: Task created content: application/json: schema: type: object properties: id: type: string description: >- Asynchronous task ID. Poll the query endpoint with this value. model: type: string description: Public model family name used for this task. required: - id - model example: id: 0f4c9f6e-1111-4c1d-8c2f-2b7b9a1c4d55 model: seedance-2.5 '400': description: >- Invalid request, for example an unsupported frames parameter or mixed content scenarios. content: application/json: schema: type: object properties: error: type: object properties: code: type: string description: Ark-style error code. message: type: string description: Client-safe error message. '401': description: Authentication failed. '402': description: Insufficient quota. '404': description: Model not found. security: - bearerAuth: [] components: securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: APIPod API key description: Use your APIPod API key as a Bearer token in the Authorization header. ````