> ## Documentation Index > Fetch the complete documentation index at: https://docs.evolink.ai/llms.txt > Use this file to discover all available pages before exploring further. # Wan2.7 Reference Video > - WAN2.7 (wan2.7-reference-video) model supports reference-to-video generation, using people or objects as protagonists to produce single-character performances or multi-character interactions - Multimodal inputs: starting frame (`image_start`), multiple reference images (`image_urls`), multiple reference videos (`video_urls`), and per-character voice bindings - **At least one** reference image (`image_urls`) or reference video (`video_urls`) must be provided; passing only `image_start` does not satisfy this. The total of `image_urls` + `video_urls` must be ≤ 5 - **Character indexing in prompt:** in Chinese use "图1, 图2 / 视频1, 视频2"; in English use "Image 1", "Video 1" — these correspond 1-based to the order of `image_urls` / `video_urls`. Images and videos are counted independently, so "Image 1" and "Video 1" can coexist - **Multi-character voice binding:** prefer `model_params.voice_bindings` (precise binding); the legacy `audio_urls` (positional alignment) is also supported - Asynchronous processing mode, use the returned task ID to [query status](/en/api-manual/task-management/get-task-detail) - Generated video links are valid for 24 hours, please save them promptly - **Billing:** charged based on "input video duration + output video duration"; only successful generations are billed, failed tasks are free ## OpenAPI ````yaml /en/api-manual/video-series/wan2.7/wan2.7-reference-video.json POST /v1/videos/generations openapi: 3.1.0 info: title: wan2.7-reference-video API description: >- Generate videos using the WAN2.7 model with reference images, reference videos, and audio inputs license: name: MIT version: 1.0.0 servers: - url: https://api.evolink.ai description: Production Environment security: - bearerAuth: [] paths: /v1/videos/generations: post: tags: - Video Generation summary: wan2.7-reference-video API description: >- - WAN2.7 (wan2.7-reference-video) model supports reference-to-video generation, using people or objects as protagonists to produce single-character performances or multi-character interactions - Multimodal inputs: starting frame (`image_start`), multiple reference images (`image_urls`), multiple reference videos (`video_urls`), and per-character voice bindings - **At least one** reference image (`image_urls`) or reference video (`video_urls`) must be provided; passing only `image_start` does not satisfy this. The total of `image_urls` + `video_urls` must be ≤ 5 - **Character indexing in prompt:** in Chinese use "图1, 图2 / 视频1, 视频2"; in English use "Image 1", "Video 1" — these correspond 1-based to the order of `image_urls` / `video_urls`. Images and videos are counted independently, so "Image 1" and "Video 1" can coexist - **Multi-character voice binding:** prefer `model_params.voice_bindings` (precise binding); the legacy `audio_urls` (positional alignment) is also supported - Asynchronous processing mode, use the returned task ID to [query status](/en/api-manual/task-management/get-task-detail) - Generated video links are valid for 24 hours, please save them promptly - **Billing:** charged based on "input video duration + output video duration"; only successful generations are billed, failed tasks are free operationId: createWan27ReferenceVideoGeneration requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/Wan27ReferenceVideoRequest' examples: single_reference_video: summary: Single reference video value: model: wan2.7-reference-video prompt: The character from the reference video dancing on a meadow video_urls: - https://example.com/reference.mp4 multi_subject_with_voice_bindings: summary: Multi-subject reference + precise voice binding (recommended) value: model: wan2.7-reference-video prompt: >- Image 1 holds Image 2 and plays a soft country folk song on the chair in Image 3, saying: "What lovely sunshine today" image_urls: - https://example.com/girl.jpg - https://example.com/object.png - https://example.com/chair.png model_params: voice_bindings: image1: https://example.com/girl_voice.mp3 duration: 10 multi_grid_storyboard: summary: Single reference image (multi-grid storyboard) value: model: wan2.7-reference-video prompt: >- Reference image, 3D cartoon adventure film style. Storyboard: 1. Wide shot of the fantasy forest; 2. The boy parts vines to scout; 3. The little robot scans ahead; 4. Close-up of treasure map; 5. The boy's excited face; 6. They leap over roots and venture deeper image_urls: - https://example.com/storyboard.png duration: 10 responses: '200': description: Video task created successfully content: application/json: schema: $ref: '#/components/schemas/VideoGenerationResponse' '400': description: Invalid request parameters content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: invalid_request message: Invalid request parameters type: invalid_request_error '401': description: Unauthenticated, invalid or expired token content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: unauthorized message: Invalid or expired token type: authentication_error '402': description: Insufficient quota, top-up required content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: insufficient_quota message: Insufficient quota. Please top up your account. type: insufficient_quota '403': description: No access permission content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: model_access_denied message: 'Token does not have access to model: wan2.7-reference-video' type: invalid_request_error '429': description: Rate limit exceeded content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: rate_limit_exceeded message: Too many requests, please try again later type: rate_limit_error '500': description: Internal server error content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: internal_error message: Internal server error type: api_error components: schemas: Wan27ReferenceVideoRequest: required: - model - prompt type: object properties: model: type: string description: Model name, must be `wan2.7-reference-video` enum: - wan2.7-reference-video example: wan2.7-reference-video prompt: type: string description: >- Text prompt for video generation. Supports Chinese and English; each character / letter / punctuation counts as 1, with overflow auto-truncated. Maximum length 5000 characters **Character indexing rules:** - Chinese: use "图1, 图2 / 视频1, 视频2" — corresponds 1-based to the order of `image_urls` / `video_urls` - English: use "Image 1", "Video 1" (capitalised, with a space between word and digit) - Images and videos are counted independently, so "Image 1" and "Video 1" can coexist - If only one reference image or one reference video is provided, you can simply write "the reference image" or "the reference video" **Multi-grid (storyboard) image:** when one multi-grid image is provided, describe key shots in storyboard form; the model recognises the grid layout and fills in the missing transitions maxLength: 5000 example: >- Video 1 holds Image 3 and plays a soft country folk tune on the chair in Image 4 negative_prompt: type: string description: >- Negative prompt describing what should not appear in the video. Supports both Chinese and English. Maximum length 500 characters; overflow is auto-truncated maxLength: 500 example: Blurry, low quality image_start: type: string format: uri description: >- Starting-frame image URL, used as the first frame of the generated video. **Does not count** toward the `image_urls` + `video_urls` ≤ 5 limit. **Does not accept voice binding** (the starting frame itself is not assigned a voice) **Use cases:** - Subject already appears in the starting frame: combine with reference materials to reinforce identity consistency - Subject not in the starting frame: reference materials define new subjects appearing as the video progresses **Image limits:** - Formats: JPEG, JPG, PNG (transparency not supported), BMP, WEBP - Resolution: width and height in `[240, 8000]` pixels - Aspect ratio: 1:8 ~ 8:1 - File size: up to `20MB` example: https://example.com/first_frame.jpg image_urls: type: array items: type: string format: uri description: >- Reference image URL array. Can supply subjects (people / animals / objects) or scene backgrounds; when a subject is included, each image should contain a **single** character **Quantity limits:** - `image_urls` + `video_urls` total ≤ 5 - At least one of `image_urls` / `video_urls` must be provided (passing only `image_start` is not enough) **Image limits:** - Formats: JPEG, JPG, PNG (transparency not supported), BMP, WEBP - Resolution: width and height in `[240, 8000]` pixels - Aspect ratio: 1:8 ~ 8:1 - File size: up to `20MB` example: - https://example.com/ref1.jpg - https://example.com/ref2.jpg video_urls: type: array items: type: string format: uri description: >- Reference video URL array. The video should ideally feature a subject (person / animal / object); empty or pure-background footage is discouraged. When a subject is included, each video should contain a **single** character. Audio in the video can be used as a voice reference **Quantity limits:** - `image_urls` + `video_urls` total ≤ 5 - At least one of `image_urls` / `video_urls` must be provided **Video limits:** - Formats: mp4, mov - Duration: `1 ~ 30` seconds - Resolution: width and height in `[240, 4096]` pixels - Aspect ratio: 1:8 ~ 8:1 - File size: up to `100MB` **Note:** when `video_urls` is provided, `duration` is capped at 10 seconds example: - https://example.com/reference.mp4 audio_urls: type: array items: type: string format: uri maxItems: 5 description: >- **[Compatibility field — prefer `model_params.voice_bindings`]** Reference voice URL array. Bound positionally to reference materials in this order: first match against `video_urls`, then against `image_urls` (in their array order, one-to-one). Up to 5 elements **Priority:** - When both `model_params.voice_bindings` and `audio_urls` are supplied, only `voice_bindings` is used and this field is ignored - If a video in `video_urls` carries audio and no voice binding is set for it, the original audio is used; an explicit voice binding overrides the original audio **Audio limits:** - Supported formats: `wav`, `mp3` - Duration range: `1 ~ 10` seconds - File size: up to `15MB` example: - https://example.com/voice1.mp3 - https://example.com/voice2.mp3 model_params: type: object description: Advanced parameter container (recommended) properties: voice_bindings: type: object description: >- **Precise voice binding for multiple characters (recommended).** Has priority over `audio_urls` **Binding rules:** - Keys are of the form `image{N}` or `video{N}` (1-based, no leading zeros, e.g. `image1`, `video2`) - The N in the key matches "图N / Image N" or "视频N / Video N" in the prompt and corresponds to `image_urls[N-1]` / `video_urls[N-1]` - Each value is the voice audio URL for that character - Total bindings ≤ 5 - `image_start` (starting frame) **does not accept** voice bindings - You may skip some characters (e.g. bind only image1 and image3, leaving image2 unbound) **Audio limits:** - Supported formats: `wav`, `mp3` - Duration range: `1 ~ 10` seconds - File size: up to `15MB` **Validation errors:** - Bad key format: `voice_bindings key "img1"/"image01" invalid` - Index out of range: `voice_bindings.image5 out of range` - Empty value, duplicate binding, or more than 5 bindings additionalProperties: type: string format: uri example: image1: https://example.com/voice_a.mp3 image3: https://example.com/voice_c.mp3 video1: https://example.com/voice_v.mp3 quality: type: string description: |- Video quality, defaults to `720p` **Options:** - `720p`: Standard definition, standard price, this is the default - `1080p`: High definition, higher price enum: - 720p - 1080p default: 720p example: 720p aspect_ratio: type: string description: >- Video aspect ratio, defaults to `16:9` **Behavior:** - `image_start` not provided: video is generated using the specified `aspect_ratio` - `image_start` provided: this field is **ignored**; the video uses an aspect ratio close to the starting-frame image **Output resolution per quality tier:** | Quality | 16:9 | 9:16 | 1:1 | 4:3 | 3:4 | | --- | --- | --- | --- | --- | --- | | 720p | 1280×720 | 720×1280 | 960×960 | 1104×832 | 832×1104 | | 1080p | 1920×1080 | 1080×1920 | 1440×1440 | 1648×1248 | 1248×1648 | enum: - '16:9' - '9:16' - '1:1' - '4:3' - '3:4' default: '16:9' example: '16:9' duration: type: number description: |- Video duration in seconds (integer) **Range:** - Without `video_urls`: `2 ~ 15`, default `5` - With `video_urls`: `2 ~ 10` (capped at 10 seconds) **Billing:** based on the actual generated video duration minimum: 2 maximum: 15 default: 5 example: 5 seed: type: integer description: >- Random seed, defaults to random **Notes:** - Range: `1` ~ `2147483647` - Fixing the seed reduces variation when iterating on prompts and improves reproducibility minimum: 1 maximum: 2147483647 example: 42 prompt_extend: type: boolean description: >- Whether to enable intelligent prompt rewriting. When enabled, a large model will optimize the prompt, which significantly improves results for simple or insufficiently descriptive prompts. **Note:** Default is `false`. Omitting the field or sending `false` will not trigger rewriting; explicitly send `true` to enable. default: false example: false callback_url: type: string description: >- HTTPS callback URL for task completion **Callback Timing:** - Triggered when task is completed, failed, or cancelled - Sent after billing confirmation **Security Restrictions:** - Only HTTPS protocol is supported - Callbacks to internal IP addresses are prohibited (127.0.0.1, 10.x.x.x, 172.16-31.x.x, 192.168.x.x, etc.) - URL length must not exceed `2048` characters **Callback Mechanism:** - Timeout: `10` seconds - Up to `3` retries after failure (retries at `1`/`2`/`4` seconds after failure) - Callback response format is consistent with the task query API response - 2xx status codes are considered successful, other status codes trigger retries format: uri example: https://your-domain.com/webhooks/video-task-completed VideoGenerationResponse: type: object properties: created: type: integer description: Task creation timestamp example: 1757169743 id: type: string description: Task ID example: task-unified-1757169743-7cvnl5zw model: type: string description: Actual model name used example: wan2.7-reference-video object: type: string enum: - video.generation.task description: Specific task type progress: type: integer description: Task progress percentage (0-100) minimum: 0 maximum: 100 example: 0 status: type: string description: Task status enum: - pending - processing - completed - failed example: pending task_info: $ref: '#/components/schemas/VideoTaskInfo' description: Detailed video task information type: type: string enum: - text - image - audio - video description: Task output type example: video usage: $ref: '#/components/schemas/Usage' description: Usage and billing information ErrorResponse: type: object properties: error: type: object properties: code: type: string description: Error code identifier message: type: string description: Error description type: type: string description: Error type VideoTaskInfo: type: object properties: can_cancel: type: boolean description: Whether the task can be cancelled example: true estimated_time: type: integer description: Estimated completion time (seconds) minimum: 0 example: 120 Usage: type: object description: Usage and billing information properties: billing_rule: type: string description: Billing rule enum: - per_call - per_token - per_second example: per_call credits_reserved: type: number description: Estimated credits to be consumed minimum: 0 example: 5 user_group: type: string description: User group category enum: - default - vip example: default securitySchemes: bearerAuth: type: http scheme: bearer description: >- ## All APIs require Bearer Token authentication ## **Get your API Key:** Visit the [API Key management page](https://evolink.ai/dashboard/keys) to obtain your API Key **Add to request headers:** ``` Authorization: Bearer YOUR_API_KEY ``` ````