WAN 3.0 Reference-to-Video

Alibaba / WAN 3.0 Reference-to-Video — Reference to Video (wan-3.0-r2v)

WAN 3.0 Reference-to-Video creates consistent videos from multimodal references — images, video clips, and audio — while maintaining appearance and style across the output.

亮點

• •Reference images – Provide images for character/object consistency.
• •Reference videos – Provide clips for motion and style reference.
• •Reference audio – Add audio for voice or sound consistency.
• •Long-form generation – Videos from 4 up to 30 seconds.

參數

• •prompt* – Text description to guide video generation. Reference inputs with "Image 1", "Video 1" labels.
• •reference_images – Reference images
• •reference_videos – Reference videos
• •reference_voice_url – Audio for voice/sound reference
• •resolution – Output video resolution
  • • 480p
  • • 720p
  • • 1080p
• •duration – Video duration in seconds (4-30)
• •enable_audio – Generate the video with an audio track
• •enable_prompt_expansion – Enable smart prompt rewriting for better quality

定價

如何使用

1. 1.Provide reference images and/or videos as URL arrays.
2. 2.Use labels like "Image 1", "Video 1" in the prompt to reference specific inputs.
3. 3.Set resolution (480p-1080p) and duration (4-30s), then submit.

專業技巧

• •Clear, front-facing reference images produce the best consistency.
• •Photorealistic real-person references may be rejected by the provider's content checks.
• •Use reference_voice_url for voice tone consistency.

更多可試用的模型