HeyOz HeyOz
Get Started
INTRODUCING WAN 2.6 BY ALIBABA

Wan 2.6

Film one clip of your spokesperson, then cast them into new scenes. Wan 2.6 takes a short reference video and carries that person's appearance and voice into generations up to 15 seconds long, at 720p or 1080p, and it can generate audio and lip-sync alongside the picture.

What people are generating

Spokesperson spots, product demos, and multi-shot social ads made with Wan 2.6.

Start creating for free

How To Use Wan 2.6 AI Video Generator

Turn one reference clip into a run of ads with Alibaba Wan 2.6 in three steps.

PICK A MODEL
STEP 1

PICK A MODEL

To get started, select the Wan 2.6 AI video model as your starting point.

PROMPT/ADD REFERENCE
STEP 2

PROMPT/ADD REFERENCE

Write your prompt, and add a reference if you have one. Wan 2.6 accepts a still image to animate, or a 2 to 30 second video of the person you want cast in the scene. You can supply your own audio track as well, or let the model generate the sound.

GENERATE AND ADJUST
STEP 3

GENERATE AND ADJUST

Watch it back, then tune the prompt, the length, or the framing and run the next variant.

Key Features of Wan 2.6

Wan 2.6 on HeyOz turns a single reference clip of your founder or presenter into a library of ad variants — same face, same voice, new scripts and new scenes. Sound and lip-sync can come out of the same generation as the picture, and one clip can hold several shots.

REFERENCE-TO-VIDEO

Cast Your Spokesperson From One Clip

Upload a 2 to 30 second video of your founder, presenter, or a creator you've booked, and Wan 2.6 uses it as a reference for both appearance and voice — then places that person in scenes you describe. Alibaba calls it China's first reference-to-video generation model. Film once with the person's agreement, then write new scripts instead of booking new shoot days.

Try it Now
AUDIO & LIP-SYNC

Sound In The Same Pass

Wan 2.6 can generate dialogue, sound effects, and ambience natively with the video, with lip-synced delivery. Bring your own audio track if you have one, or let the model dub the line — either way the sound is handled in the same step rather than a separate pass afterwards.

Try it Now
MULTI-SHOT

Several Shots, One Generation

Wan 2.6 handles shot transitions itself, working to hold your character and lighting steady across the cuts. Fifteen seconds is enough room for a hook, a demo, and a call to action — the standard paid-social slot — without stitching clips together afterwards.

Try it Now
FORMATS

720p And 1080p, Shaped For The Feed

Generate at 720p for volume testing or 1080p when you're shipping the winner — up to 1920x1080. Aspect ratios cover 16:9 for YouTube and display, 1:1 for feed, plus 4:3.

Try it Now

Professional videos for every use case

Built around one idea: shoot your talent once, with their agreement, then keep the campaign running from the keyboard.

Founder UGC At Volume

Record one reference clip of your founder, then generate a week of angles in their face and voice. New script, new hook, no new shoot.

Social Ad Variants

Generate at 1080p ready to post to Meta, TikTok and the rest, with the option to have sound generated in the same pass.

Multi-Shot Product Spots

Fit a hook, a product demo, and a call to action into one 15-second generation, with the cuts handled inside the model.

Ads From A Product Still

Start from a single photo of the SKU and let image-to-video put it in motion — on a counter, in a hand, under studio light.

Localized Versions

Re-voice a spot for another market and the model can return lip-synced delivery, rather than subtitles pasted over the original cut.

Frequently Asked Questions

What is Wan 2.6?

Wan 2.6 is Alibaba's video model from the Tongyi Wanxiang family, released on December 16, 2025. Alongside text-to-video and image-to-video it adds reference-to-video: give it a short clip of a person and it uses that clip as the guide for how they look and sound in the scenes you write. It can also generate audio natively with the picture. On HeyOz it's wired into your brand kit and ad formats.

How do I use Wan 2.6 on HeyOz?

Choose Wan 2.6 as your model, describe the scene, and attach a reference if you're working from an image or a clip of your presenter. Generate, watch it back, adjust, and run the next variant.

What is reference-to-video?

You upload a 2 to 30 second video of a person, and Wan 2.6 draws on it for both their appearance and their voice, then generates new scenes with them in it. Alibaba describes it as China's first reference-to-video generation model. Practically, it means one filmed clip of your spokesperson can carry a whole run of ad variants.

Can I use a reference video of any person?

No — use footage of someone who has agreed to it. Reference-to-video carries a real person's face and voice into new material, so the clip should be of your founder, your team, or talent you've booked with rights for this use. Uploading a video of someone who hasn't consented isn't something to run an ad on, and it isn't something HeyOz is for.

How long can the videos be?

Wan 2.6 supports generations from 2 to 15 seconds, with 5 seconds as the default. The available lengths differ between Alibaba's endpoints, so check the duration options in the studio before you plan around a 15-second cut.

What resolution can Wan 2.6 output?

There's a 720p tier and a 1080p tier — 1920x1080 landscape, 1440x1440 square, plus 1632x1248. Draft on the lower tier while you're testing angles and generate at 1080p for the cut you intend to run.

What inputs does Wan 2.6 accept?

A prompt of up to 1,500 characters and a negative prompt of up to 500. On top of that: an image in JPEG, PNG, BMP, or WEBP between 360 and 2000 pixels and under 25MB; a reference video of 2 to 30 seconds; and an optional audio track in WAV or MP3, 3 to 30 seconds and under 15MB.

Is Wan 2.6 open source?

No. Alibaba published earlier Wan releases under Apache 2.0 with downloadable weights, but 2.6 is served through the API only — there's no public checkpoint or repository for it. Some articles say otherwise; they're describing the older versions. Note too that Alibaba's API reference also documents Wan 2.7, so 2.6 is available but no longer the newest release in the family.

Every leading model, on HeyOz

Pick the right engine for the job — all in one place.

Seedance 2.0 Seedance 2.5 Veo 3.1 Google Omni Flash Seedance 1.5 Pro Kling 3 Kling O3 Kling O3 Pro GPT-Image-2 Nano Banana 2 Nano Banana Pro

Why HeyOz

Anyone can give you Wan 2.6. Only HeyOz turns it into your ad.

On-brand by default

HeyOz reads your brand and keeps every asset on-color, on-voice, on-message.

Every model, one place

Seedance, Veo, Kling, GPT Image, and more — no juggling subscriptions.

Ad-ready output

Sized and formatted for Meta and TikTok, straight out of the box.

A whole workflow

Not just a model — templates, avatars, and an agent that does it all.

Every ad your brand needs, in one place

Start Now. No agency, no brief, no blank screen.

Start creating now