Wan 3.0

What is Wan 3.0?#

Wan 3.0 is Alibaba's video generation model, live on Unifically in two variants: alibaba/wan-3.0-video and the faster alibaba/wan-3.0-video-prime. It generates 2 to 30 seconds of video with audio in a single pass, double the 15-second ceiling of Wan 2.7.

The bigger change is what it takes as input. Beyond text and images, reference mode accepts up to 10 reference images, 5 reference videos, and 5 reference audio clips in one request, plus a document (PDF, DOCX, PPTX, XLSX) or a web page URL. Upload a product deck, add a prompt, and the output is a finished promotional video.

Audio generates with the picture: voices, sound design, and ambience arrive in the same pass, and a single audio parameter turns it off when you want a silent video.

Key features of Wan 3.0#

• Up to 30 seconds in one generation. Room for complete camera moves and unbroken shots instead of joining short videos. Set duration anywhere from 2 to 30.
• Documents and web pages as input. A PDF, DOCX, PPTX, or XLSX through file_url, or a page URL through link_url, parses directly into video.
• Up to 20 reference assets. One request carries 10 images, 5 videos, and 5 audio clips, enough for a full set of product shots, style frames, and sound cues.
• Pixel-accurate reference consistency. Characters, props, layouts, styles, and audio reproduce from references with strict detail control, not rough resemblance.
• Native audio in every mode. Voices and sound design generate with the frames, on by default and switchable off per request.
• A fast Prime variant. Wan 3.0 Prime (alibaba/wan-3.0-video-prime) takes the same inputs and returns results noticeably faster for time-sensitive work.

Best for#

Use cases#

Turn a product deck into a launch video: the slides carry the facts, the prompt sets the tone, and the model returns one finished piece with voice-over. Build 30-second ads as one continuous shot instead of stitching three generations. Convert documentation and reports into training videos for teams that will not read a PDF. For short dramas and social stories, the longer timeline plus micro-expression rendering keeps a performance running through multiple beats, and reference mode holds the same character across every scene you generate.

Limitations#

Reference inputs, documents, and links live in reference mode only; image-to-video takes a first frame or a first + last frame pair and nothing else. A text prompt is required in text and reference modes. Arena lists no Wan 3.0 entry on its video boards as of August 25, 2026, so there is no independent rank for it yet, only Alibaba's own material. And unlike Wan 2.1 and 2.2, no open weights are announced for this release.

Wan 3.0 vs Wan 2.7#

Wan 3.0 doubles the single-pass ceiling to 30 seconds against Wan 2.7's 15, adds 480P output for cheap drafts, and takes input types 2.7 does not have: documents, web pages, reference audio, and a 20-asset reference budget. Wan 2.7 keeps two abilities 3.0 does not expose here: video continuation from an input video and a separate edit route (alibaba/wan-2.7-video-edit). Use Wan 3.0 for long-form, document-driven, and reference-heavy work; stay on Wan 2.7 when you need to continue or edit an existing video.

What Wan 3.0 can do#

API examples#

Call Wan 3.0 from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/alibaba/wan-3.0-video.

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "alibaba/wan-3.0-video",
    "input": {
      "mode": "t2v",
      "prompt": "A kitten running in the moonlight, cinematic lighting",
      "resolution": "1080P",
      "ratio": "16:9",
      "duration": 10
    }
  }'

Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.