Providers and ModelsAlibaba Cloud

Wan 3.0

Wan 3.0
Alibaba Cloud · Video Generation
POST /v1/videos/generations

All-in-one video generation up to 30 seconds with native audio, omni-modal references (images, video, audio, files, web links), and a fast Prime tier.

At a glance

FieldValue
Model idwan-3-0
Model release date2026-08-06
Input modalitiesText, Image, Video, Audio
Output modalitiesVideo
Context window-
Weight precision-
RegionSingapore
Featurestext_to_video, image_to_video, reference_to_video, audio_sync, character_consistency
Native inferenceNo
NewYes
Supported endpointsPOST /v1/videos/generations
Alternate model idsalibaba/wan-3.0, wan-3.0, wan3.0, wan3.0-video, wan3.0-video-prime

Pricing

ChargeSpecRate
Standard 480Pper second$0.07
Standard 720Pper second$0.14
Standard 1080Pper second$0.28
Prime 480Pper second$0.136
Prime 720Pper second$0.28
Prime 1080Pper second$0.56

Example request

$curl https://api.empiriolabs.ai/v1/videos/generations \
>  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
>  -H 'Content-Type: application/json' \
>  -d '{"model": "wan-3-0", "prompt": "sunrise over the ocean", "duration": 6}'

Parameters

ParameterTypeRequiredDefaultDescription
promptstringyes-Scene description, up to 20,000 characters. In reference mode, address inputs in your prompt as Image 1, Video 1, Audio 1.
model_tierenumno"standard"standard: full quality at the base rate. prime: significantly faster end-to-end generation at a higher per-second rate. Each tier is billed at its own rate. · Allowed: standard, prime
modeenumno"auto"Auto detects from attachments. t2v: text-to-video. i2v: animate one image, or interpolate between two (first and last frame). r2v: omni reference from images, videos, audio, a document, or a web link. · Allowed: auto, t2v, i2v, r2v
resolutionenumno"1080p"Output resolution. Each resolution bills at its own per-second rate. · Allowed: 480p, 720p, 1080p
durationnumberno5Clip length in seconds. Pass -1 via the API to let the model choose the length. With video references, input plus output stays within 30 seconds. · Range: 2 – 30
aspect_ratioenumno"adaptive"adaptive recommends an aspect ratio from your inputs and prompt. · Allowed: adaptive, 16:9, 9:16, 1:1, 4:3, 3:4
imagestringno-Image URL. i2v uses one or two images (first and last frame); r2v accepts up to 10 references.
videostringno-Reference video URL (r2v). Up to 5 clips, 15 seconds combined.
seednumberno-Reproducibility seed. · Range: 0 – 2147483647
audiobooleannotrueGenerate the soundtrack with the video. Turning it off does not change the price.
watermarkbooleannofalseAdd a provenance watermark to the output.
prompt_extendbooleannotrueLet the service rewrite the prompt for better results.
linkstringno-Public web page URL used as a reference (r2v). Provide either a link or a file_url, not both.
file_urlstringno-Document URL used as a reference (r2v): PDF, Office, or text files up to 100 MB and 50 pages. Provide either a file_url or a link, not both.

Notes

Generation can take several minutes on the standard tier; the prime tier is significantly faster end to end. Modes are auto-detected from attachments, or override with the mode parameter. The model_tier parameter picks standard or prime; each tier bills at its own per-second rate.

Modes

  • T2V: no attachments
  • I2V: 1 image (first frame), or exactly 2 images (first + last frame)
  • R2V (omni reference): up to 10 images, 5 videos (15s combined), and 5 audio clips (15s combined), plus 1 document file or 1 public web link; address them in your prompt as Image 1, Video 1, Audio 1
  • I2V frames and R2V references cannot be combined in one request

Inputs

  • Images: JPEG, PNG, BMP, or WEBP, 240 to 8,000 px per side, up to 20 MB
  • Videos: MP4 or MOV, 1-15s per clip, up to 100 MB
  • Audio: WAV or MP3, 1-15s per clip, up to 15 MB

Duration and billing

  • Clips run 2-30 seconds; pass duration: -1 to let the model choose the length
  • With video references, input plus output share the 30-second budget, and billed seconds cover input + output combined
  • Output is MP4 at 30 fps with native audio; audio: false disables sound without changing the price

Uploaded media preprocessing

  • Reference videos are normalized to provider-compatible MP4 when needed.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/wan-3-0.