Wan 2.6

byAlibaba (Tongyi Lab / Wan team)

Direct a polished, cinematic mini-scene with cuts, spoken dialogue, foley and a music bed, all from a single prompt

Roll the scene

How Wan-2.6 Video works

Write the scene like a shot list. Timed brackets set the cuts, quoted lines set the voices, and optional images, tracks or reference clips set the look and the sound.

Multi-shot direction prepared for Wan 2.6 Video.

Structure your multi-shot timeline

Write an overall scene, then timed shots like 'First shot [0-5s]' with camera moves, transitions and quoted dialogue.

Opening frame connected to Wan 2.6 Video.

Link reference assets and audio

Add a first-frame image, a music track, or up to three reference videos tagged @Video1 to @Video3 in your prompt.

Wan 2.6 Video node with compact input, controls, and generated output.

Render the synced sequence

Pick 5, 10 or 15 seconds at 720p or 1080p, then generate cuts, lip-synced speech and sound in one pass.

What Wan-2.6 Video is good at

Cuts that keep their continuity, mouths that follow the words, and identity that survives a new scene: this is built for short narrative clips rather than single-shot loops.

Multi-shot timeline scheduling

Plan several cuts inside one generation, up to 15 seconds. Character, lighting and scene details hold across wide, medium and close framings. Make your timestamps add up to the chosen duration.

Native dialogue lip-sync

Speech, effects and music are generated together with mouths that track the words. Put lines in quotes and name who speaks so voices attach to the right character.

Reference-video character carry-over

Tag up to three clips as @Video1 to @Video3 to bring a person, animal or object into new scenes with the same look and voice. The reference-to-video variant is capped around 10 seconds and ignores the aspect ratio setting.

First-frame image animation

The image-to-video variant treats your still as the opening frame, so product shots and key frames from an image model keep their composition. Later shots can drift, and the crop follows the image's aspect ratio.

Made with Wan-2.6 Video

A gallery of 5 to 15 second clips: macro liquid, gliding interiors, a spoken cut, a music-driven monolith and a footwear study, each with sound included.

Macro product reveal with dynamic fluid interaction

Slow interior camera tracking with shifting geometric light

Multi-shot narrative cut with speech synchronization

Surreal music visualizer framed on a solitary basalt pillar

Footwear motion study against high-contrast studio darkness

What people build with Wan-2.6 Video

Ad makers, filmmakers, product designers and musicians reach for it when a finished short needs picture, speech and score together.

Social ads with sound baked in

01

Build vertical spots with cuts, a spoken hook and a music bed in one render, so a finished 9:16 draft arrives without a separate audio pass.

Brand commercial and product spots

02

Animate a hero product still, or let the model invent the look from words. Iterate at 720p, then commit to 1080p for the final cut.

Short film and scene previs

03

Block out a micro-story with wide, medium and close framings and speaking characters before committing to a full shoot.

Product and industrial design reveals

04

Show a device, vehicle or piece of gear in motion from a still keyframe, with lighting and materials held steady through the move.

Music-driven visualizers and posters

05

Feed in your own track as the background music and let imagery move to it for lyric loops, single-cover motion and teaser clips.

Wan-2.6 Video specs

Plans and credits
Inputs
Prompt (text), Image Prompt (image), Video References (up to 3 videos), Audio, Negative Prompt (text)
Output
video
Duration
5s, 10s, 15s
Resolution
720p, 1080p
Aspect ratios
16:9, 9:16, 1:1, 4:3, 3:4
Cost
689–3,100 credits per run

Wan 2.6 questions

They differ in what anchors the clip. Text-to-video starts from words alone, for concept scenes, multi-shot stories and ads, but character and product looks vary between seeds. Image-to-video uses your still as the first frame, though later shots can drift from it. Reference-to-video carries a person, animal or object from your footage into new scenes.

Use reference-to-video. Upload up to three clips and tag them @Video1 to @Video3 in the prompt, and the subject's look and voice carry into the new scene. Output is capped around 10 seconds, the aspect ratio setting is ignored, and clear, well-lit references give cleaner identity.

Up to 15 seconds, in 5, 10 or 15 second options, at 720p or 1080p. It is the wrong choice for longer continuous footage. When using multi-shot, make your timed shots add up to the duration you select.

Yes. Dialogue, effects and music are generated in the same pass, with lip-sync that tracks the speech. Put lines in quotes and say who speaks. You can also upload a track as background music, but it is truncated to the clip length, and it is likely ignored when reference videos supply the voice.

Exact on-screen text and logos can garble, and acrobatic or physically demanding human motion can look floaty or stiff. Voices can drift with many speakers or long dialogue. Turn multi-shot off for a single continuous take, or unwanted cuts can appear.

Wan 2.6 comes from Alibaba's Tongyi Lab Wan team and was released in December 2025. It is a commercial, hosted-only model, so it is not suited to fully local or open-weight workflows.

Try Wan-2.6 Video on Fuser

Direct a polished, cinematic mini-scene with cuts, spoken dialogue, foley and a music bed, all from a single prompt

Roll the scene

Related models

Wan-2.6 Video docsAll models

MiniMax H3 Recast

MiniMax

Kling O3

Kuaishou (Kling AI)

Kling O3 Edit

Kuaishou Technology

Kling 3.0 Video

Kuaishou

Seedance 2

ByteDance (Seed team)

MiniMax H3

MiniMax

Wan-2.6 Video in our guides

All articles