Vidu Video

The Vidu video series currently provides three integration methods: the official format (/vidu/ent/v2/*), OpenAI-compatible /v1/videos (multipart/form-data format), and the unified video endpoints /v1/video/create and /v1/video/query (JSON).

Official Format

Use the /vidu/ent/v2/* series of interfaces, which support multiple video generation modes.

Text to Video

POST /vidu/ent/v2/text2video, generates video from text only.

Image to Video

POST /vidu/ent/v2/img2video, generates video from a first-frame image.

Start-End Frame to Video

POST /vidu/ent/v2/start-end2video, generates video from start and end frames.

Reference to Video

POST /vidu/ent/v2/reference2video, generates video with subject reference.

Query Task

GET /vidu/ent/v2/task/{task_id}/creations, queries task status and results.

Supported Models (Official Format)

ModelDescriptionDuration Rules
viduq3-proEfficiently generates high-quality audio and video contentDepends on the model
viduq2 / viduq2-proNew model with good results and rich details1-10 seconds
viduq2-turboNew model with good results and fast generation1-10 seconds
viduq1 / viduq1-classicClear visuals and smooth transitionsDepends on the model
vidu2.0Fast generation speedDepends on the model

OpenAI Format

Use POST /v1/videos to submit and GET /v1/videos/{task_id} to query, with submission in multipart/form-data.

Video Generation

POST /v1/videos, submits a Vidu video task.

Task Status Query

GET /v1/videos/{task_id}, queries task status and results.

Supported Models (OpenAI Format)

ModelDescription
viduq3-proEfficiently generates high-quality audio and video content, making the video more vivid, more expressive, and more immersive
viduq2-proNew model with good results and rich details
viduq2-turboNew model with good results and fast generation

Unified Video

Use POST /v1/video/create to create tasks, and GET /v1/video/query?id= for queries.

Generate Video

POST /v1/video/create, submits a Vidu task in the unified video format.

Query Task

GET /v1/video/query, queries status and results by task id.

Common Capabilities (Unified Video)

CapabilityParameter PassingDescription
Text to Videoimages: []Text prompt only
First-Frame to Videoimages 1 imageFirst-frame image + text prompt
Start-End Frame to Videoimages 2 imagesFirst frame + end frame
Reference to Videosubjects arraySubject reference images

Comparison

ItemOfficial FormatOpenAI FormatUnified Video
Create/vidu/ent/v2/*POST /v1/videosPOST /v1/video/create
QueryGET /vidu/ent/v2/task/{task_id}/creationsGET /v1/videos/{task_id}GET /v1/video/query?id=
Request FormatJSONmultipart/form-dataJSON
Reference Imagesimages / subjectsimagesimages
Aspect Ratioaspect_ratioaspect_ratioaspect_ratio
Resolutionresolutionresolutionsize
Durationdurationdurationduration

Parameter Constraints

FieldAvailable ValuesDescription
aspect_ratio16:9, 9:16, 3:4, 4:3, 1:13:4 and 4:3 are supported only by q2 and q3 models
resolution540p, 720p, 1080pResolution specification
duration1-10Video duration (seconds), q2 defaults to 5
stylegeneral, animeStyle (not effective for q2, q3)
FeatureDescription
Audio and Video Outputaudio: true, outputs videos with dialogue and sound effects (q3 only)
Background Musicbgm: true, automatically adds BGM
Off-Peak Modeoff_peak: true, generates during off-peak times for lower credits
Recommended Promptsis_rec: true, the system automatically recommends prompts
Watermarkwatermark: true, adds a fixed watermark