wan-3.0/reference-to-video

Key Features

• Multimodal reference inputs — combine first/last frame images with additional reference images, videos, and audio in a single request.
• Motion and camera reference — extract choreography and camera movement from a reference video and apply it to a new scene.
• Style and character consistency — reference images preserve appearance and style across the generated clip.
• Native audio-video synthesis — synchronized dialogue, music and sound effects generated automatically alongside the picture.
• Native 30-second clips — orchestrate longer, multi-scene sequences from a single reference set.
• Bilingual prompt support — supports both English and Chinese prompts natively.

Parameters

How to Use

1. Upload reference assets — first/last frame images anchor the start and end, additional images define style/character, videos define motion/camera, audio drives soundtrack.
2. Write a prompt describing how the references should come together into a scene.
3. Choose duration (up to 30s), resolution, and aspect ratio.
4. Generate — submit and wait, then download the finished video with synchronized audio.
5. Iterate — start with 480p/5s for quick iterations, then render the final version at 1080p.

Code Examples

import os
import requests

response = requests.post(
    "https://aircube.ai/api/v3/wan-3.0/reference-to-video",
    headers={
        "Authorization": "Bearer " + os.environ["AIRCUBE_API_KEY"],
        "Content-Type": "application/json",
    },
    json={
    "prompt": "A dancer performs the same choreography as the reference video, but on a neon-lit rooftop at night",
    "image": "https://example.com/dancer-first-frame.jpg",
    "reference_videos": [
        "https://example.com/choreography-reference.mp4"
    ],
    "duration": 8,
    "resolution": "1080p",
    "aspect_ratio": "16:9"
},
    timeout=300,
)
data = response.json()

if data["success"]:
    print("ID:", data["data"]["id"], "Status:", data["data"]["status"])
else:
    print("Error:", data["error"]["message"])

Pricing

Billing rules

• 480p / 4s: $0.20.
• 720p / 4s: $0.40.
• 1080p / 4s: $0.80.
• Longer durations scale proportionally.
• First 5 reference images are free; each additional image adds $0.04.
• Reference video input is billed per second at $0.05/s (480p), $0.10/s (720p), $0.20/s (1080p), added on top of the output price.
• Failed generations are not charged.

Best Use Cases

• Motion transfer — extract choreography or camera movement from a reference video and apply it to new subjects.
• Style transfer — apply a reference image's look to the generated scene.
• Multi-scene narrative — combine multiple references to maintain character and camera continuity across a 30-second cut.
• Precise start/end control — pin the exact first and last frame while the model fills in the motion between them.
• Soundtrack-guided generation — use a reference audio clip to drive the pacing and mood of the scene.

Pro Tips

• Combine a first-frame and last-frame image whenever the clip needs to land on a specific pose or composition.
• Keep reference videos short and focused — Wan 3.0 reads up to 15 seconds combined across all reference videos.
• Describe each reference's role explicitly in the prompt to avoid ambiguity.
• Use one primary camera instruction per prompt for the cleanest motion transfer.
• Start with 480p/5s for rapid iteration, then render the final version at 1080p.

Notes

• Up to 10 reference images, 5 reference videos (15s combined), and 5 reference audio clips (15s combined) per request.
• Native audio generation is enabled by default.
• Duration range: 2-30 seconds (continuous selection).
• Successor to Wan 2.7 Reference-to-Video with a longer duration ceiling and first/last-frame control.