← Node reference

WAN 3.0

AI generationAlibaba

Alibaba WAN 3.0. One model for text, frames, reference images, reference clips and reference audio. Up to 30 seconds with native sound.

Runs
On our servers
Typical time
60-300s
Inputs
22
Outputs
1

Inputs

PortTypeID
Prompttextprompt-in
Reference 1optionalimageimage-in
Last Frameoptionalimageimage-2
Video 1optionalvideovideo-in
Video 2optionalvideovideo-2
Video 3optionalvideovideo-3
Video 4optionalvideovideo-4
Video 5optionalvideovideo-5
Audio 1optionalaudioaudio-in
Audio 2optionalaudioaudio-2
Audio 3optionalaudioaudio-3
Audio 4optionalaudioaudio-4
Audio 5optionalaudioaudio-5
Reference 2optionalimageimage-3
Reference 3optionalimageimage-4
Reference 4optionalimageimage-5
Reference 5optionalimageimage-6
Reference 6optionalimageimage-7
Reference 7optionalimageimage-8
Reference 8optionalimageimage-9
Reference 9optionalimageimage-10
Reference 10optionalimageimage-11

Outputs

PortTypeID
Videovideovideo-out

Settings

SettingTypeDefaultOptions
Duration (seconds)With a clip wired, input plus output must stay under 30s.slider52 to 30
ResolutionOutput resolutionselect720P480p, 720p, 1080p
Aspect RatioAuto takes its shape from what you wire inselectadaptiveAuto, 16:9 Landscape, 9:16 Portrait, 1:1 Square, 4:3 Standard, 3:4 Portrait
SoundGenerate a native audio trackselectonOn, Off
Reference ModeHow your wired images are read. Reference clips and audio work with either.selectframesStart / end frames, Reference images
SeedFixed number for reproducible results, 0 = randomnumber00 to 2147483647

Prompting

Describe the shot. Reference media is addressed as Image1, Video1, Audio1...

Maximum prompt length: 20000 characters.

What this connects into (20)

Computed from the port compatibility rules, so it stays correct as nodes change.

Node ID wan-30-video. Open it in Studio with Tab, then type the name.