Endpoints and what they accept
| Endpoint | Use when | Required inputs |
|---|---|---|
| text-to-video | Pure prompt, no visual input | prompt, aspect_ratio, resolution, duration |
| image-to-video | Animate a starting frame | prompt, first_frame, aspect_ratio, resolution, duration |
| reference-to-video | Hold characters and props consistent | prompt, up to 20 mixed-modality references, duration |
| document-to-video | Convert a slide deck or PDF to a video | document_reference, prompt, duration |
| video-edit | Edit an existing clip | prompt, source_video, edits, duration |
| video-extend | Continue an existing clip | prompt, source_video, duration |
Minimal request
curl -X POST "https://api.oneinfer.ai/v1/ula/generate-video" \
-H "Authorization: Bearer $ONEINFER_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"model": "alibaba/wan-3.0-prime",
"prompt": "A quiet city street at golden hour as camera slowly pans up to the skyline",
"aspect_ratio": "16:9",
"resolution": "720p",
"duration": 5
}'Parameters dropped in Prime
Wan 3.0 Prime removes the negative_prompt and prompt_extend parameters that standard Wan 3.0 exposes. Code written against standard Wan 3.0 needs those parameters removed before it will run on Prime.
Model identifier variants
Three distinct identifiers resolve to the same checkpoint: alibaba/wan3.0-video-prime on Alibaba Model Studio, alibaba/wan-3.0-prime on OpenRouter, and wan-v3.0-video-prime on Vercel AI Gateway. All three spell the same Wan 3.0 Prime model.
Ready to test the workflow?
Create account & add creditsAudio is on by default at no extra cost
Audio is generated alongside the picture in a single pass. Disable with audio=false; the price does not change either way.
Upstream rate limits
Alibaba limits the upstream to 2 concurrent requests, a 50 task async queue and 30 requests per minute. Route through an aggregator such as OneInfer to run higher concurrency against the model.
Frequently asked questions
How do I use Wan 3.0 Prime with the API?
Send a POST to /v1/ula/generate-video with model "alibaba/wan-3.0-prime", a valid API key, and a prompt plus aspect_ratio, resolution and duration. Add first_frame, reference items or document_reference only when you need them.
How many references can Wan 3.0 Prime accept?
Up to 20 mixed-modality references per request — up to 10 images, 5 video clips and 5 audio clips. That is the highest reference image budget of any routed video model on OneInfer.
Does Wan 3.0 Prime generate audio in the same pass?
Yes. Audio is composed alongside the picture in one asynchronous call. Disable with audio=false; the price does not change either way.
Does Wan 3.0 Prime support negative prompts?
No. Prime removes the negative_prompt and prompt_extend parameters that standard Wan 3.0 exposes. Remove those parameters before submitting to Prime.
Where is the canonical Wan 3.0 Prime page on OneInfer?
The canonical model page with current pricing, providers, and capabilities is /models/alibaba/wan-3.0-prime. This page is a focused facet of that entity, not a replacement for it.
How should I treat benchmark or price claims on Prime?
Check each claim's provenance label and observed date. Vendor-reported, independently verified and derived numbers are shown as separate evidence classes on this hub.
How do I call Wan 3.0 Prime from the API?
POST to /v1/ula/generate-video with model "alibaba/wan-3.0-prime", a valid API key, and a prompt plus aspect_ratio and resolution. Reference items and document inputs are optional and are added as separate payload fields.
Put Wan 3.0 to work
Fund a controlled evaluation, send a reference frame or document, and measure quality and cost on your own workload.