> ## Documentation Index > Fetch the complete documentation index at: https://docs.jimmyai.cn/llms.txt > Use this file to discover all available pages before exploring further. # Create Gemini Omni Video > 创建 Gemini Omni 视频生成任务。任务创建后，请使用通用视频任务查询接口获取生成状态和结果。 After creation, the response returns a `task_id`. Use the [Query Video Task](/en/api-reference/common/query) endpoint to poll task status and read the final video URL from `result.video_url`. ## Supported models | model | Notes | | -------------- | --------------------------------------------------------------- | | `Gemini-Omini` | Standard Gemini Omni — 4/6/8/10s with 720p/1080p/4k | | `omni-10s` | Fixed 10s, 720p, up to 7 reference images (image or video URLs) | Aliases such as `Omni-Flash-Ext` and `gemini-omni` normalize to `Gemini-Omini`. ## Reference Video & Generation Modes The standard version of `Gemini-Omini` supports the following advanced parameters to enable video reference and custom modes: * `generation_type` (string): Generation mode, supporting `frame` (first-frame mode, where `image_urls` can only take 1 image as the first frame of the video) or `reference` (reference mode, where `image_urls` can accept 1 or 3 reference images). Defaults to `reference`. * `video_urls` (array): A list of reference video URLs (currently supports up to 1 video). **Note: `duration` is not required when uploading a reference video, and `duration` and `video_urls` cannot be passed at the same time.** ## omni-10s example ```bash theme={null} curl -X POST 'https://www.jimmyai.cn/api/open-api/v1/gemini/omni/videos' \ -H 'Authorization: Bearer sk_xxx' \ -H 'Content-Type: application/json' \ -d '{ "model": "omni-10s", "prompt": "Character from the reference runs through a cyberpunk street, cinematic tracking shot", "aspect_ratio": "16:9", "image_urls": ["https://example.com/ref1.jpg"] }' ``` For `omni-10s`, `duration` and `resolution` in the request are ignored; billing uses the `omni-10s` model config (10s, 720p). ## OpenAPI ````yaml POST /api/open-api/v1/gemini/omni/videos openapi: 3.1.0 info: title: Jimmy AI OpenAPI description: API for Jimmy AI video generation services version: 1.0.0 servers: - url: https://www.jimmyai.cn description: Production server security: - bearerAuth: [] paths: /api/open-api/v1/gemini/omni/videos: post: summary: Create Gemini Omni Video Task description: >- Create a Gemini Omni video generation task. After the task is created, use the common video task query endpoint to retrieve status and results. operationId: createGeminiOmniVideo requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/GeminiOmniGenerationRequest' responses: '200': description: Task created successfully content: application/json: schema: $ref: '#/components/schemas/VideoGenerationResponse' components: schemas: GeminiOmniGenerationRequest: type: object required: - model - prompt properties: model: type: string enum: - Gemini-Omini - Omni-Flash-Ext - gemini-omni - omni-10s description: >- Model name. Gemini-Omini, Omni-Flash-Ext, and gemini-omni normalize to Gemini-Omini; omni-10s is a fixed 10s model (720p, up to 7 reference images). prompt: type: string description: Text prompt for video generation duration: type: integer default: 6 description: >- Video duration in seconds. Default: 6. Only supports: 4, 6, 8, 10. Note: duration is not required when uploading a reference video, and duration and video_urls cannot be passed at the same time. resolution: type: string default: 720p enum: - 720p - 1080p - 4k description: Video resolution. Defaults to 720p. aspect_ratio: type: string description: >- Video aspect ratio, for example 16:9 or 9:16. Takes precedence over orientation. orientation: type: string enum: - landscape - portrait description: >- Video orientation. Used only when aspect_ratio is empty: landscape maps to 16:9, portrait maps to 9:16. image_urls: type: array items: type: string description: Reference image URL list. omni-10s supports up to 7 images. images: type: array items: type: string description: >- Compatibility field for common OpenAPI image inputs; used when image_urls is empty. omni-10s supports up to 7 images. generation_type: type: string default: reference enum: - frame - reference description: >- Generation type. frame: first frame mode (image_urls can only accept 1 image as the first frame of the video); reference: reference mode (image_urls can accept 1 or 3 images as style/subject reference). Defaults to reference. video_urls: type: array items: type: string description: >- Reference video URL list (currently supports up to 1 video). Note: duration is not required when uploading a reference video, and duration and video_urls cannot be passed at the same time. example: model: Gemini-Omini prompt: >- A cat running through a neon rainy street, cinematic low-angle tracking shot duration: 6 resolution: 720p aspect_ratio: '16:9' generation_type: reference image_urls: [] video_urls: [] VideoGenerationResponse: type: object properties: code: type: string example: '20000' msg: type: string example: ok data: type: object properties: task_id: type: string status: type: string model: type: string created_at: type: integer format: int64 securitySchemes: bearerAuth: type: http scheme: bearer ````