Long single takes
Build a 30-second scene as one continuous take, with sound generated alongside the picture.
Try workflow
Wan 3.0
Turn a description, a deck or a webpage into 30 seconds of finished video. Direct it with invideo's agents. They write the prompts.
Trusted by teams at
Why creatives choose Wan 3.0
30 seconds in one continuous take
One generation runs up to 30 seconds, double Wan 2.7's 15. A scene with a setup, a build and a payoff fits in one take. No cuts you didn't ask for.
Builds video from documents and webpages
Give it a PDF, deck, spreadsheet or URL and Wan 3.0 turns the content into video. Your reports, sales decks and training material are already the script.
Smart duration
The model fits the clip length to the prompt. No padding, no beats cut short.
Up to 20 references per generation
Ten images, five video clips and five audio clips, plus a document or webpage. Video and audio references can run 15 seconds each. A face, a product shot, a gesture and a voice all land in the same shot.
Section editing
Select part of a finished clip and regenerate only that part, dialogue included. The rest of the take stays as approved.
Multiple angles in one generation
Get a wide, a close-up and a reverse in a single output, with the location consistent across cuts. The sequence arrives assembled.
One model, four modes
Text-to-video, image-to-video, first-and-last-frame and reference-to-video, all in one model. Wan 2.7 needed four.
Try Wan 3.0How Wan 3.0 works in invideo?
On invideo, Wan 3.0 runs inside an agentic workflow. The agent picks it for the right shots, writes the prompts, carries your references and edits sections instead of re-rolling clips. You approve every step.
The agent picks the right model.
Wan 3.0 is one of 200+ models. The agent sends it the work it does best: long single takes, video built from documents or webpages, and shots with many references. You only pick a model if you want to, and your own routing rules hold for the whole project.
You direct, the agent writes the prompts.
Describe the shot in plain language. The agent turns it into the structured prompt a 30-second take needs, so you never learn the model's quirks.
Your documents become source material, not homework.
Drop a deck, report or webpage into the project. The agent passes it straight to Wan 3.0, so you never have to summarise it into a prompt.
Your references carry into every generation.
Approve your characters, voice and look once. The agent saves them to the project's memory and attaches them to every Wan 3.0 call. Nothing gets re-uploaded at scene forty.
Your notes become section edits, not regenerations.
Tell the agent the middle of a take isn't working. It regenerates only that section, and the rest of your approved take stays intact.
You always stay in control.
Choose how much the agent does alone: generate everything, ask before each video, or show you every prompt first. Change it any time.
What Wan 3.0 does best
invideo's agents handle the prompting. You direct.
Complete ad spots
Generate a full spot in one pass. If a section isn't converting, edit that section, not the whole ad.
Try workflow
Pitch decks into video
Turn a client deck into a video version of the same pitch. No rewriting it as a prompt.
Try workflow
Explainers and internal comms
Build explainers straight from the handbook, process doc or training deck your team already wrote.
Try workflow
Wan 3.0 vs Wan 2.7
Feature | Wan 3.0 | Wan 2.7 |
|---|---|---|
Max clip length | 30 seconds | 15 seconds |
Smart duration | Yes | No |
Document and webpage input | Yes | No |
Max references per generation | Up to 20 | Up to 5 (video), 9 (image) |
Native audio | Yes | Yes |
Section editing | Regenerate a chosen section | Whole clip, prompt-based |
First and last frame control | Yes | Yes |
Max video resolution | 1080p | 1080p |
Unified model | One model | Four models |
Helping creatives stay creative
Multiplayer mode
Collaborate in real time with live cursors to show what everyone's working on.
RRebeccajust now
Can we push the train arrival 2 sec later? Feels rushed.
AAiko2m
Love the warmth here — keep this lighting for the reunion shot.
Storyboarding
Turn any script or idea into a shot-by-shot plan, then tweak as needed before generating.
Script writing
Write your script inside invideo, and ask an AI co-writer for help if you'd like.
Timeline editor
Picture Premiere Pro with full AI.
Build your own agents
Create custom agents to fill specific roles like cinematographer, music designer, and more.
From solo creatives to creative enterprises
Private & SecureSOC 2 & ISO AlignedGDPR Compliant
World-class investors stand behind invideo.
Backed by the firms behind Stripe, Spotify, Flipkart, and ByteDance.
Loading pricing…
Wan 3.0 FAQs
What is Wan 3.0 and what can it do?
Wan 3.0 is Alibaba's video model. It generates up to 30 seconds with sound in one pass and takes text, images, video, audio, documents and URLs as input. It supports first-and-last-frame control and can regenerate one section of a clip. On invideo, you direct in plain language and the agent writes the prompts.
How long can a Wan 3.0 video be?
Up to 30 seconds per generation, double Wan 2.7. Smart duration fits the length to the prompt instead of padding to a fixed clip.
Can Wan 3.0 make video from a PDF or deck?
Yes. It reads documents and webpage URLs as native inputs and generates video from the content.
How many references can Wan 3.0 use?
Up to 20: ten images, five video clips and five audio clips, plus one document or webpage. Video and audio references can each total 15 seconds.
Does Wan 3.0 generate audio?
Yes. Audio is generated in the same pass, so clips come back with sound.
What resolution does Wan 3.0 support?
Up to 1080p.
How does Wan 3.0 compare to Wan 2.7?
Wan 3.0 doubles clip length, adds document and webpage input, smart duration and section editing, raises the reference limit, and merges four video modes into one model. It's video-only, so use Wan 2.7 for stills.
Do I need to learn Wan 3.0's prompting to use it on invideo?
No. The agent writes the prompts in the structure Wan 3.0 needs, references and documents included. Describe the shot the way you'd brief a person.
Is Wan 3.0 open source?
No. Wan 3.0 is closed-weight and hosted only. Alibaba's last open-weight video flagship is Wan 2.2.
Is Wan 3.0 free on invideo?
Wan 3.0 is billed at its API rate, like every generative model on invideo. Regular use needs an active plan with credits. Credits reset monthly and don't roll over.