muapi Logomuapi Logo
muapi Logomuapi Logo

Wan 3.0 Reference to Video: Image-to-Video

Generate video guided by reference images, video, and audio with Wan 3.0. Keep characters and style consistent across a scene. Pay per generation. Wan 3.0 Reference to Video generates a video guided by a prompt plus up to 10 reference images, 5 reference videos, and 5 reference audios for visual and audio coherence.

Interactive model controls

Wan 3.0 Reference to Video generates a video guided by a prompt plus up to 10 reference images, 5 reference videos, and 5 reference audios for visual and audio coherence.

📝

Overview

About this model

Wan 3.0 Reference to Video generates a video guided by a text prompt plus up to 10 reference images, 5 reference videos, and 5 reference audios, letting you keep characters, style, and sound consistent across a generated scene. Reference media are identified by their order in each array, so the prompt can address "the first reference image" or "the person in the reference video" directly. Choose 480p, 720p, or 1080p output, five aspect ratios, and a duration from 2 to 30 seconds; when reference videos are supplied, their total duration plus the generated output must stay within 30 seconds. For a plain text-to-video generation with no reference media, see Wan 3.0 Text to Video; for animating a single source image, see Wan 3.0 Image to Video. For the higher-fidelity Prime tier, see Wan 3.0 Prime Reference to Video. For a bolder, higher-contrast take on the same request shape, see Wan 3.0 Spicy Reference to Video.

1Character consistency: Carry a character's appearance across a new generated scene using reference images.
2Style transfer: Apply the visual style of a reference video to a new prompt-driven shot.
3Soundtrack matching: Sync a generated video to an existing reference audio track.
4Multi-reference composition: Combine several reference images and a reference video into one coherent scene.
💰

Pricing & Value

Cost analysis

muapiappFrom $0.10 per generation (720p, 2s) up to $6.00 (1080p, 30s)

Pay-per-generation, priced by resolution and duration, no subscription required.

Fal.aiNot available

Wan 3.0 is not listed on Fal.ai at this time.

ReplicateNot available

Wan 3.0 is not listed on Replicate at this time.

** Competitor pricing is estimated based on similar model architectures and usage tiers.

⚙️

Technical Details

Configuration schema

Promptstring

Describe the video you want to create. Reference media are identified by their array order.

Default ValueThe two people from the reference images meet on the rainy street shown in the reference video.
Reference Image URLsarray

Up to 10 reference images for visual coherence guidance.

Default Valueundefined
Reference Video URLsarray

Up to 5 reference video clips (MP4/MOV, 1-15s each); total reference + generated duration must not exceed 30s.

Default Valueundefined
Reference Audio URLsarray

Up to 5 reference audio clips (total duration up to 15s) for soundtrack synchronization.

Default Valueundefined
ResolutionEnum (3 options)

Output video resolution.

Default Value720p
Aspect RatioEnum (6 options)

Output video frame dimensions.

Default Value16:9
Durationinteger

Video length in seconds.

Default Value5
Thinking Modeboolean

Enable deep-thinking mode for complex scene understanding.

Default Valuefalse
Enable Audioboolean

Include a generated audio track with the video.

Default Valuetrue
Seedinteger

Random seed for reproducibility. Use -1 for a random seed.

Default Value-1
📖

Implementation Guide

Developer documentation

How to Use Wan 3.0 Reference to Video

  1. Write the prompt: Describe the scene, referring to reference media by their array order (e.g. "the person from the first reference image").

  2. Attach reference media: Provide up to 10 URLs in images_list, up to 5 in videos_list, and up to 5 in audios_list.

  3. Set the output shape: Choose resolution (480p, 720p, 1080p), aspect_ratio, and duration (2-30 seconds, capped so reference-video duration plus output stays at or under 30 seconds).

  4. Tune generation: Enable thinking_mode for complex multi-reference scenes, and set enable_audio to false for a silent video.

  5. Submit and poll: Send the request with your Muapi API key, store the returned request_id, then poll the prediction result endpoint or supply a webhook_url.

curl -X POST https://api.muapi.ai/api/v1/wan3.0-reference-to-video \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "The person from the reference image walks into the room shown in the reference video.",
    "images_list": ["https://example.com/character.jpg"],
    "videos_list": ["https://example.com/room.mp4"],
    "resolution": "720p",
    "duration": 5
  }'
❓

Common Questions

Frequently asked

What does Wan 3.0 Reference to Video do?

It generates a video guided by a prompt plus reference images, videos, and audio, keeping characters, style, and sound consistent with the supplied references.

How many reference files can I send?

Up to 10 reference images, 5 reference videos, and 5 reference audios per request.

How does the model know which reference is which?

Reference media are identified by their order in each array, so the prompt can refer to "the first reference image" or similar ordinal phrasing.

Is there a duration limit when using reference videos?

Yes. When reference videos are supplied, their total input duration plus the generated output duration must not exceed 30 seconds.

What resolutions does Wan 3.0 Reference to Video support?

480p, 720p, and 1080p, with five aspect ratios (16:9, 9:16, 1:1, 4:3, 3:4).

How is Wan 3.0 Reference to Video billed?

Pricing scales with resolution and duration, and the completed task response includes the exact charge for that generation.