mingyan
Videos - Create Task
Create video generation tasks using Videos models with support for text-only, first/last frame, and image/video references; audio references are supported by selected models
Videos Create Video Task
Create video generation tasks using the Videos model family. This API supports
videos, videos_stable, and videos_stable_fast, including text-to-video, first/last frame control, reference images, reference videos, and audio references. videos supports up to 3 audio references; videos_stable and videos_stable_fast support up to 1 audio reference.Authentication
Get KeyAll requests require a Bearer token in the request header:
cURL
Authorization: Bearer {{key}}
Base URL
https://zcbservice.aizfw.cn/kyyReactApiServerbaseUrl is the shared prefix for all public APIs. The api field in the current page frontmatter shows the full endpoint. Use this baseUrl as the common prefix when reading or composing request paths.Model Overview
Supported models:
videos- Videos standard model, billed per request, supports wider aspect ratios and audio referencesvideos_stable- Video2 full model, billed per requestvideos_stable_fast- Video2 fast model, billed per request
Request Parameters
modelstringrequiredModel name.
Available values:
videos- Videos standard model, supports audio referencesvideos_stable- Video2 full modelvideos_stable_fast- Video2 fast model
promptstringrequiredVideo generation prompt.
videos_stable / videos_stable_fast support up to 5000 characters; videos is not limited to 5000 characters.Avoid prohibited, infringing, political, or explicit content.
Example:
"A cute kitten playing in the grass"durationintegerrequiredVideo duration in seconds. Supported range:
4-15.ratiostringOutput aspect ratio. Defaults to
16:9.Supported values:
videos:21:9,16:9,4:3,1:1,3:4,9:16videos_stable/videos_stable_fast:16:9,9:16,1:1
resolutionstringOutput resolution. Defaults to
720p.Currently supported:
720p
First/Last Frame Mode
first_imagestringFirst frame image URL.
- Only used for first/last frame scenarios
- Must be used together with
last_image
Cannot be used together with
referenceImages or referenceVideos; first/last frame mode does not support referenceAudioslast_imagestringLast frame image URL.
- Must be used together with
first_image
Cannot be used together with
referenceImages or referenceVideos; first/last frame mode does not support referenceAudiosCompatibility notes:
- The legacy fields
imageandlastFrameImageare still supported - When both new and legacy fields are provided,
first_image/last_imagetake priority - If both new and legacy fields are provided with different values, the API returns a parameter conflict
Reference Material Mode
referenceImagesarrayReference image URL list.
Rules:
videossupports up to9imagesvideos_stable/videos_stable_fastsupport up to4images- Each image must not exceed
20MB - Can be combined with
referenceVideosfor image/video guidance - Can be combined with
referenceAudios;videossupports up to 3 audios, whilevideos_stable/videos_stable_fastsupport up to 1 audio
Cannot be used together with
image or lastFrameImagereferenceVideosarrayReference video URL list.
Rules:
- Up to
3videos - Total duration must not exceed
15seconds - Total size must not exceed
200MB - Each video must have a resolution between
720pxand2160px
Cannot be used together with
image or lastFrameImagereferenceAudiosarrayReference audio URL list.
Rules:
videossupports up to3audiosvideos_stable/videos_stable_fastsupport up to1audio- Total duration must not exceed
15seconds - Used to provide audio style, rhythm, or sound references for video generation
- Can be combined with
referenceImagesandreferenceVideosfor multimodal guidance
Passing more audio URLs than the selected model supports returns an error; first/last frame mode does not support audio references
Response Parameters
idstringUnique identifier for the video generation task, used for subsequent status queries.
objectstringObject type, always
video.createdintegerTask creation timestamp.
modelstringModel name used for generation.
statusstringTask status:
queued- Queuedprocessing- Processingcompleted- Completedfailed- Failed
errorstringError message, returned when the task is
failed.Parameter Selection Rules
Mutually Exclusive Rules:
- First/last frame mode:
first_imageandlast_imagemust be used together - Reference material mode:
referenceImagesandreferenceVideoscan be used individually or together;referenceAudiossupports up to 3 audios withvideos, and up to 1 audio withvideos_stable/videos_stable_fast - Mode exclusivity: First/last frame mode cannot be combined with reference material mode or audio references
Use Cases
Text to Video
Use only
prompt, model, duration, and other base parametersFirst/Last Frame
Use
first_image and last_image for precise start and end frame controlMultimodal References
Combine
referenceImages, referenceVideos, and referenceAudios for stronger generation guidanceModel Comparison
| Model | Billing | Duration Range | Reference Support | Aspect Ratios |
|---|---|---|---|---|
videos | Per request | 4-15 seconds | Up to 9 images, 3 videos, and 3 audios | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 |
videos_stable | Per request | 4-15 seconds | Up to 4 images, 3 videos, and 1 audio | 16:9, 9:16, 1:1 |
videos_stable_fast | Per request | 4-15 seconds | Up to 4 images, 3 videos, and 1 audio | 16:9, 9:16, 1:1 |
Best Practices:
- Video generation is asynchronous, so save the returned
idfor later queries. - Include scene, subject, motion, camera, and style details in the prompt.
- In first/last frame mode, make sure the two images are visually continuous.
- When using video or audio references, prefer short, clear, and subject-focused source materials.
- For the Video2 fast model, use
videos_stable_fast.
Prompt Suggestions:
- Describe the subject, action, and camera movement
- Add style keywords such as "cinematic", "realistic", or "dreamlike"
- If you need camera motion or scene transitions, describe them explicitly

