Wan 3.0 Reference-to-Video is built on Alibaba's Wan 3.0 all-in-one video model (wan3.0-video). Combine up to 10 reference images, 5 reference videos and 5 reference audios, and refer to them as Video 1 / Image 1 in the prompt. With thinking mode enabled, it can also parse documents and web pages. Output is natively up to 30 seconds with smart duration; when reference videos are used, input video duration plus output duration must be ≤ 30s. Supports 480P/720P/1080P, adaptive aspect ratio, and an optional audio track.
Multimodal Reference (Image/Text/Video/Audio)Pixel-Level ConsistencyNative Up to 30sSmart Duration480P/720P/1080PThinking Mode for Docs/WebControllable Audio Output
image-to-video 7.9 k
Playground
Input
*API Key
Please select API Key
*prompt
videoUrls
imageUrls
audioUrls
Upload Image
Supports JPG/PNG format, file size no more than 10MB