Wan 3.0 Prime

Wan 3.0 Prime text to video uses the model id alibaba/wan-3.0-prime/text-to-video. Wan 3.0's top tier — one take up to 30 s at 1080p, 10 images and 5 clips

1. Calling the API

Endpoint

Send a POST to https://api.elyum.ai/v1/generations with a JSON body matching the parameters below. Every model uses the same endpoint; the model field picks this one.

const res = await fetch("https://api.elyum.ai/v1/generations", {  method: "POST",  headers: {    Authorization: `Bearer ${process.env.ELYUM_API_KEY}`,    "Content-Type": "application/json",    Prefer: "wait=55",  },  body: JSON.stringify({    "model": "alibaba/wan-3.0-prime/text-to-video",    "prompt": "A slow push-in on a lighthouse at dusk, waves breaking below"  }),});const generation = await res.json();console.log(generation.status, generation.outputs);

Request parameters

A value outside these lists is refused with a 400 naming the field, before anything is charged. Nothing is quietly rounded or dropped.

2. Authentication

API key

Create a key on Keys with the generate and spend scopes and send it as a bearer token. Keep it on your server; never send it to a browser.

Authorization: Bearer $ELYUM_API_KEY

3. Queue

Submit and wait

With Prefer: wait=55 the request holds until the render ends or 55 seconds pass. A finished render answers 200 with its outputs; one still running answers 202 with its id.

Poll and cancel

const headers = { Authorization: `Bearer ${process.env.ELYUM_API_KEY}` }; // Poll until the render ends: succeeded, failed or canceledlet generation = await (await fetch(`https://api.elyum.ai/v1/generations/${id}`, { headers })).json();while (generation.status === "queued" || generation.status === "running") {  await new Promise((r) => setTimeout(r, 5000));  generation = await (await fetch(`https://api.elyum.ai/v1/generations/${id}`, { headers })).json();} // Or stop it (a hold is refunded)await fetch(`https://api.elyum.ai/v1/generations/${id}/cancel`, { method: "POST", headers });

Output links are signed and last 24 hours: download them to your own storage, or fetch the generation again for fresh ones.

4. Pricing

A struck price is the list price before today's. Send ?dry_run=true for the exact price of a request with nothing held. The API spends your API balance; a render that fails or is canceled costs nothing.