wan3.0-video Alibaba Wan 3 unifies text, frame, multimodal-reference, audio-driven, editing, and extension video in one 2–30 second model.
No GPU. No setup. Cancel anytime.
Wan 3 is Alibaba Cloud Model Studio's unified premium video model on Sogni. Alibaba currently labels the upstream API as preview (as of August 2026). One model ID covers text-to-video, first-frame and first-and-last-frame animation, loose multimodal references, audio-driven generation, source-video editing, and source-video extension.
It renders 2–30 second clips at fixed 30 fps in 480p, 720p, or 1080p. Choose a fixed 16:9, 4:3, 1:1, 3:4, or 9:16 canvas, use adaptive shape, or let smart duration select the timing. Native audio is on by default and can be disabled without changing the rate.
Wan 3 accepts up to 10 images, 5 reference videos, and 5 reference audio clips. It can also use one public document up to 100 MB or one public webpage as context; PDF, DOCX, DOC, PPTX, PPT, Keynote, and Pages inputs are limited to 50 pages. Document and webpage context are mutually exclusive and cannot be combined with dedicated first/last-frame anchors.
Wan 3 accepts plain natural language and optional provider expansion. Make each visible beat and each reference role explicit.
Wan 3 is a pay-as-you-go Premium Spark vendor model. Every clip uses purchased Spark, including for Sogni Unlimited members. Unlimited saves 5%; Unlimited Pro saves 10%.
| Configuration | Spark | USD |
|---|---|---|
| 5 s · 480p | 65.0 Spark | $0.33 |
| 5 s · 720p | 130 Spark | $0.65 |
| 5 s · 1080p | 260 Spark | $1.30 |
| 30 s · 480p | 390 Spark | $1.95 |
| 30 s · 720p | 780 Spark | $3.90 |
| 30 s · 1080p | 1560 Spark | $7.80 |
Sogni retail rates before plan discounts (as of 2026-08): $0.065 per output second at 480p, $0.13 at 720p, and $0.26 at 1080p. Native audio on or off uses the same rate.
Smart duration reserves the 30-second maximum before generation; settlement uses Alibaba's reported output duration after the job completes.
One Sogni API key reaches every model on the Supernet — call Wan 3 with the exact model id.
import { SogniClient } from '@sogni-ai/sogni-client';
const client = await SogniClient.createInstance({
appId: crypto.randomUUID(),
apiKey: process.env.SOGNI_API_KEY,
network: 'fast',
});
const project = await client.projects.create({
type: 'video',
modelId: 'wan3.0-video',
positivePrompt: 'A documentary presenter walks through a detailed workshop and says Welcome to the future as the camera tracks beside her',
numberOfMedia: 1,
duration: 8,
ratio: '16:9',
wan3TaskType: 'create',
generateAudio: true,
promptExtend: true,
watermark: false,
});
const [url] = await project.waitForCompletion();
console.log(url); // result link — download within 24hcurl https://api.sogni.ai/v1/creative-agent/workflows \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $SOGNI_API_KEY" \
-d '{
"input": {
"title": "First render",
"steps": [{
"id": "step1",
"toolName": "generate_video",
"arguments": { "prompt": "A documentary presenter walks through a detailed workshop and says Welcome to the future as the camera tracks beside her", "videoModel": "wan3.0-video", "duration": 5 }
}]
},
"confirm_cost": true
}' REST workflow alias: wan3.0-video. Pin the exact id with the SDK. Full reference at docs.sogni.ai.
A 30-second source dance becomes a branded reference-video edit: Mark keeps his blue tux and identity while Sogni's broad, plush pink sloth-unicorn mascot performs beside him. Because Wan 3 caps source duration plus output duration at 30 seconds, the edit uses two 15-second R2V passes joined under the original continuous soundtrack.
Wan 3 is a third-party frontier model. It uses purchased Spark for every clip; Unlimited plans discount those pay-as-you-go rates instead of covering them as fair-use renders.
Vendor-backed renders are billed with Spark even when your account has an active Unlimited subscription.
Unlimited saves 5%; Unlimited Pro saves 10%.
Launch Wan 3 from the Sogni app or call the exact model id through the API.
Use premium vendor models alongside Sogni Supernet image, video, music, and language models.
Use text-to-video for a prompt alone; image-to-video for a first frame or first-and-last-frame pair; reference generation for loose image, video, audio, document, or webpage context; and video-to-video for source-video editing or extension. All use wan3.0-video.
Yes. Native audio is enabled by default and can be disabled for silent output. Audio on and off have the same per-second price.
Alibaba prompt expansion defaults on for direct Wan 3 requests. Turn it off when wording must remain exact. When Sogni Chat has already shaped a brief, it disables the provider's second expansion so the prompt is not rewritten twice.
Yes. Both operations require a source video. Extension also requires adaptive ratio and a prompt that clearly asks to continue the supplied clip; source-video duration plus requested output cannot exceed 30 seconds.
Yes. Supply one public HTTPS document up to 100 MB or one public HTTPS webpage. PDF, DOCX, DOC, PPTX, PPT, Keynote, and Pages inputs are limited to 50 pages. The two options are mutually exclusive, and neither can be combined with dedicated first/last-frame anchors.
Wan 3 is pay-as-you-go Premium Spark. It is not covered as an Unlimited fair-use render; every clip uses purchased Spark. Unlimited saves 5%; Unlimited Pro saves 10%. Prices start at 65.0 Spark ($0.33) per clip.
Create in the app, or build with the API. Your call.