POST /v2/videos/generations

Method: POSTEndpoint: /v2/videos/generations

The Tongyi Wanxiang text-to-video model generates a smooth video from a text prompt. Supported capabilities include:

Core capabilities: flexible durations (5s/10s), specified video resolution (480P/720P/1080P), smart prompt rewriting, and watermark support.

Audio capabilities: supports automatic dubbing or a custom audio file for audio-video synchronization. Available only on wan2.5.

Request Parameters

Header Parameters

Authorization
string
Optional
Default Value:
Bearer {{YOUR_API_KEY}}

Body Parameters application/json

prompt
string
Required
The text prompt supports Chinese and English, with a maximum length of 800 characters. Each Chinese character or letter counts as one character. Content that exceeds this limit will be truncated.
Text prompts support both Chinese and English, with a maximum length of 800 characters. Each Chinese character or letter counts as one character. Content beyond this limit is truncated.
Example: A kitten running in the moonlight.
Example: a kitten runs under the moonlight.
model
enum<string>
Required
Model name. Example: wan2.1-t2v-turbo.
Enum Values:
wan2.2-t2v-plus
Latest models
  // provider-specific example normalized for English documentation
wanx2.1-t2v-turbo
  // provider-specific example normalized for English documentation
  // provider-specific example normalized for English documentation
wanx2.1-t2v-plus
  // provider-specific example normalized for English documentation
  // provider-specific example normalized for English documentation
duration
integer
Optional
Duration of the generated video in seconds. This parameter is fixed at 5 and cannot be changed. The model always generates a 5-second video.
size
string
Optional
  // provider-specific example normalized for English documentation
  // provider-specific example normalized for English documentation
  // provider-specific example normalized for English documentation
480P: The optional video resolutions and their corresponding aspect ratios are:
  // provider-specific example normalized for English documentation
832×480: 16:9.  832×480：16：9。
480×832: 9:16.  480×832：9：16。
624×624: 1:1.  624×624：1。
720P: The optional video resolutions and their corresponding aspect ratios are:
  // provider-specific example normalized for English documentation
1280×720: 16:9.  1280×720：16：9。
720×1280: 9:16.  720×1280：9：16。
960×960: 1:1.  960×960：1：1。
1088×832: 4:3.  1088×832：4：3。
832×1088: 3:4.  832×1088：3：4。
1080P: The optional video resolutions and their corresponding aspect ratios are:
  // provider-specific example normalized for English documentation
1920×1080: 16:9.
1920×1080：16：9。
1080×1920: 9:16.
1080×1920：9：16。
1440×1440: 1:1.  1440×1440：1：1。
1632×1248: 4:3.  1632×1248：4：3。
1248×1632: 3:4.  1248×1632：3：4。
watermark
boolean
Optional
Specifies whether to add a watermark. The watermark appears in the lower-right corner and reads "Generated by AI".
template
string
Optional
negative_prompt
string
Optional
A negative prompt is used to describe content that you do not want to appear in the video, which lets you restrict the video content.
The negative prompt describes content you do not want to appear in the video, helping you constrain the result.
It supports Chinese and English, with a maximum length of 500 characters. Content that exceeds this limit will be truncated.
It supports both Chinese and English, with a maximum length of 500 characters. Content beyond that limit is truncated.
Examples: low resolution, error, worst quality, low quality, defects, extra fingers, poor proportions.
Example: low resolution, errors, worst quality, low quality, defects, extra fingers, bad proportions.
prompt_extend
boolean
Optional
Specifies whether prompt rewriting is enabled. When enabled, a large language model (LLM) intelligently rewrites the input prompt. This significantly improves results for shorter prompts but increases processing time.
seed
integer
Optional
A random seed used to control the randomness of the generated content. The value must be in the range [0, 2147483647].
If this parameter is not provided, the algorithm automatically generates a random seed. To keep the generated content relatively stable, reuse the same seed value.
Example
{
"prompt"
:
"dance"
,
"model"
:
"wan2.2-t2v-plus"
}

Example Request

Shell

curl --location --request POST '/v2/videos/generations' \
--header 'Authorization: Bearer {{YOUR_API_KEY}}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "prompt": "dance",
    "model": "wan2.2-t2v-plus"
}'

Response

🟢 200 Success

Content Type: application/json

Response Schema

task_id
string
Required

Example

{
    "task_id": "e7bed961-d1b9-4b3f-8ef9-5f441bde28c8"
}