Wan 3.0 Reference to Video
Generate video from up to 10 reference images with Alibaba's Wan 3.0. The model carries the people, objects, and styles from your references into a continuous take of 2 to 30 seconds, with native synchronized audio generated in the same pass. Output at 480p, 720p, or 1080p — reference images are free, and audio adds nothing to the price.
20% less than the priciest provider (5 compared). Same model, same output.
Up to 10 reference images contributing people, objects, or styles. Cite in the prompt as Image 1, Image 2, ... At least one reference (image, video or audio) is required.
Up to 5 reference clips (mp4/mov, 1-15s each, 15s combined) for a subject's look and voice, motion, choreography or camera work. Cite as Video 1, ... Their length is billed as extra seconds, and reference + output must stay within 30s.
Up to 5 audio clips (wav/mp3, 1-15s each, 15s combined): a voice or song to lip-sync, dance or time the motion to. Cite as Audio 1, ... Free.
Resolution
Output resolution. 480p is the budget rung; 1080p is the premium rung.
Aspect Ratio
Aspect ratio of the generated video. 'adaptive' lets the model choose the best framing.
Duration
Length of the generated video in seconds (2-30). Billed per second. Wan 3.0 generates up to 30 seconds in a single take.
Audio
Generate synchronized audio (dialogue, effects, music) in the same pass. Same price either way; some sellers cannot switch audio off, so turning it off can change the winner.
20% less than the priciest provider (5 compared). Same model, same output.
Your result renders here.
Use Wan 3.0 Reference to Video via the API
Generate with Wan 3.0 Reference to Video from your own code. Async by design: submit a generation, get a run id back instantly, poll until it completes. Billed in credits at the best available provider price — from $0.025 per second.
1 — Start a generation
curl -X POST https://app.scenetra.com/api/v1/generate/wan-3.0-ref2v \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"prompt":"The person from the first image walks through the street from the second, handheld follow shot, crowd murmur","reference_image_urls":["https://example.com/person.jpg","https://example.com/street.jpg"],"duration":"5"}'Response (202)
{
"runId": "run_abc123",
"status": "PENDING",
"model": "wan-3.0-ref2v",
"estimatedCredits": 400,
"statusUrl": "/api/v1/runs/run_abc123"
}2 — Poll until COMPLETED (every 2-5 seconds)
curl https://app.scenetra.com/api/v1/runs/run_abc123 \
-H "Authorization: Bearer YOUR_API_KEY"Response
{
"runId": "run_abc123",
"status": "COMPLETED",
"progress": { "completed": 1, "total": 1 },
"outputs": [
{ "type": "video", "url": "https://images.scenetra.com/..." }
],
"error": null,
"estimatedCredits": 400
}Request fields
| Field | Type | Description | Default |
|---|---|---|---|
| prompt* | string | Describe the video: subject, motion, camera work, and audio direction. Native synchronized audio is generated in the same pass. | — |
| reference_image_urls | string[] (https URLs) | Up to 10 reference images contributing people, objects, or styles. Cite in the prompt as Image 1, Image 2, ... At least one reference (image, video or audio) is required. | — |
| reference_video_urls | string[] (https URLs) | Up to 5 reference clips (mp4/mov, 1-15s each, 15s combined) for a subject's look and voice, motion, choreography or camera work. Cite as Video 1, ... Their length is billed as extra seconds, and reference + output must stay within 30s. | — |
| reference_audio_urls | string[] (https URLs) | Up to 5 audio clips (wav/mp3, 1-15s each, 15s combined): a voice or song to lip-sync, dance or time the motion to. Cite as Audio 1, ... Free. | — |
| resolution | string | Output resolution. 480p is the budget rung; 1080p is the premium rung.480p · 720p · 1080p | 720p |
| aspect_ratio | string | Aspect ratio of the generated video. 'adaptive' lets the model choose the best framing.adaptive · 16:9 · 4:3 · 1:1 · 3:4 · 9:16 | adaptive |
| duration | string | Length of the generated video in seconds (2-30). Billed per second. Wan 3.0 generates up to 30 seconds in a single take.2 · 3 · 4 · 5 · 6 · 7 · 8 · 9 · 10 · 11 · 12 · 13 · 14 · 15 · 16 · 17 · 18 · 19 · 20 · 21 · 22 · 23 · 24 · 25 · 26 · 27 · 28 · 29 · 30 | 5 |
| audio | string | Generate synchronized audio (dialogue, effects, music) in the same pass. Same price either way; some sellers cannot switch audio off, so turning it off can change the winner.true · false | true |
| callbackUrl | string (https URL) | Optional. When the run finishes (completed or failed), we POST the final run status to this URL — same shape as the poll response, plus model, params, and chargedCredits. Best-effort; polling remains the source of truth. | |
Unknown fields are rejected with a helpful 400. Media fields take https URLs. There is no provider parameter — every run is routed to the cheapest capable provider automatically.
Good to know
Billing
Runs bill your account credits at provider cost with 0% markup. The submit response includes the credit estimate; failed generations are refunded automatically. Add ?dryRun=1 to get the estimate without running anything.
Errors
Errors are always { "error": { "code", "message" } } with stable codes: unauthorized, model_not_available, invalid_parameter, insufficient_credits, rate_limited. A failed run is not an HTTP error — poll responses report status: "FAILED" with the reason.
What makes Wan 3.0 Reference to Video different
Single takes up to 30 seconds
Most video models top out at 10 or 12 seconds per generation and force you to stitch clips together. Wan 3.0 generates any length from 2 to 30 seconds in one pass — long enough for a complete ad spot, a full scene, or an entire product walkthrough with no cuts, no extends, and no continuity drift between segments.
Native audio — and it's the cheaper setting
Wan 3.0 generates dialogue, sound effects, and music synchronized to the picture in the same pass, steered by your prompt. Unusually, leaving audio on is also the cheaper configuration: Scenetra routes every render to the cheapest provider eligible for your exact settings, and audio-on renders qualify for a lower rate — $0.025 per second at 480p versus $0.04 with audio switched off. Unless you specifically need a silent clip, keep the default.
Three resolutions, one price ladder
480p at $0.025 per second is the drafting rung — a full 30-second take costs $0.75. The 720p default runs $0.05 per second, and 1080p finals cost $0.10 (audio-on rates). A practical workflow on a Scenetra board: iterate long takes at 480p until the pacing and audio land, then flip the resolution select to 1080p for the final render — same node, same wiring.
Billed at cost in a visual workflow
Wan 3.0 runs as a node in Scenetra's node-based editor, so you can chain it with image models — generate a style frame, refine it, then feed the look into your video prompts on one canvas. Every generation is billed at provider cost with 0% markup, pay-per-generation with no subscription gate.
Parameters
| Parameter | Type | Description | Default |
|---|---|---|---|
| Prompt* | text | Describe the video: subject, motion, camera work, and audio direction. Native synchronized audio is generated in the same pass. | — |
| Resolution | select | Output resolution. 480p is the budget rung; 1080p is the premium rung.480p · 720p · 1080p | 720p |
| Aspect Ratio | select | Aspect ratio of the generated video. Auto lets the model choose the best framing for your prompt.Auto · 16:9 · 4:3 · 1:1 · 3:4 · 9:16 | Auto |
| Duration | select | Length of the generated video in seconds. Wan 3.0 generates any length from 2 to 30 seconds in a single take.2 · 3 · 4 · 5 · 6 · 7 · 8 · 9 · 10 · 11 · 12 · 13 · 14 · 15 · 16 · 17 · 18 · 19 · 20 · 21 · 22 · 23 · 24 · 25 · 26 · 27 · 28 · 29 · 30 | 5 |
| Audio | boolean | Generate synchronized audio (dialogue, SFX, music) in the same pass. Leaving it on is also the cheaper setting — see pricing. | true |
Pricing
From $0.025 per second
Pay per generation — only pay for what you use.
480p (with audio)
$0.025/s
per second
480p (audio off)
$0.04/s
per second
720p (with audio)
$0.05/s
per second
720p (audio off)
$0.08/s
per second
1080p (with audio)
$0.10/s
per second
1080p (audio off)
$0.16/s
per second
What a video costs
| Duration | 480p (audio) | 480p (muted) | 720p (audio) | 720p (muted) | 1080p (audio) | 1080p (muted) |
|---|---|---|---|---|---|---|
| 5 seconds | $0.125 | $0.20 | $0.25 | $0.40 | $0.50 | $0.80 |
| 10 seconds | $0.25 | $0.40 | $0.50 | $0.80 | $1.00 | $1.60 |
Use cases
Marketing & ads
- Complete 30-second spots in one take
- Product teasers with synced sound
- Vertical 9:16 social ads
- Multi-length campaign variants
Social content
- Sound-on shorts and reels
- Long-form single-shot storytelling
- Ambient mood loops
- Square and portrait cuts from one prompt
Film & previz
- Full-scene previz without stitching
- Dialogue and SFX scratch tracks
- Establishing shots with ambience
- Cheap 480p blocking passes
Education & explainers
- Narrated explainer segments
- Step-by-step process clips
- Concept visualizations with sound
- Course intro sequences
Frequently asked questions
What is Wan 3.0 reference-to-video?+
It's the mode of Alibaba's Wan 3.0 that generates video from up to 10 reference images. Instead of animating one starting frame, the model uses your references to keep specific people, objects, and styles consistent throughout the generated clip — 2 to 30 seconds in a single take, with native synchronized audio. On Scenetra it runs as a node in the visual workflow editor.
How many reference images can I use?+
Between 1 and 10 per generation, and they're free — you pay only the per-second video rate. You can mix reference types in one job: a character, a product, and a style frame together, with the prompt describing how they interact.
What's the difference between reference-to-video and image-to-video?+
Image-to-video animates one upload as the literal first frame of the clip. Reference-to-video accepts up to 10 images that condition the content — the people, objects, and styles they contain appear consistently anywhere in the shot, in scenes your images never showed. Use image-to-video to bring a finished composition to life, and reference-to-video to keep a cast or look consistent across many shots.
Does reference-to-video generate audio too?+
Yes — like the other Wan 3.0 modes it generates dialogue, sound effects, and music synchronized with the picture in the same pass. In this mode the price is identical with audio on or off, so there's no cost reason to disable it.
Related models
Wan 3.0 Prime
Wan 3.0 Prime is Alibaba's accelerated Wan 3.0: the same model capabilities, generated up to seven times faster. It renders up to 30 seconds of video in a single take with native synchronized audio (dialogue, sound effects and music) produced in the same pass as the picture. Pick any duration from 2 to 30 seconds at 480p, 720p or 1080p, start from a prompt, a first frame, or up to 10 reference images, videos and audio clips. Scenetra routes every render to the cheapest provider for your settings.
View model →Seedance 2.5
ByteDance's Seedance 2.5 generates cinematic videos up to 30 seconds from text prompts with synchronized audio — voice, sound effects, and background music — built in by default. The newest Seedance generation, with mp4 or high-fidelity mov output for multi-round editing pipelines.
View model →Kling 3.0 Pro
Kuaishou's flagship video generation model delivering stunning visual fidelity. Kling 3.0 Pro supports text-to-video, image-to-video with start/end frames, element references for character consistency, and native audio generation.
View model →Start creating with Wan 3.0 Reference to Video
Use Wan 3.0 Reference to Video alongside 50+ other AI models in Scenetra's visual workflow editor. No setup required.
Get Started