MILabs
Tools

Wan AI video generation model: uses and setup for small business

Wan is Alibaba's open video generation model. Learn what businesses use it for and how a self-hosted setup is arranged.

Updated 2026-09-16

What Wan is

Wan is a family of open video generation models built by Alibaba's Wan team. Its published code and model weights sit in the Wan-Video repository on GitHub, where the team documents text-to-video and image-to-video generation. A text-to-video model takes a written description and produces a short clip from it. An image-to-video model starts from a still picture and adds motion to it. Both approaches suit businesses that already hold photographs, product images or written briefs and want quick moving drafts from them. Because the model is released for download, it can run on hardware the business controls as well as through a hosted service. That makes Wan one option among several for teams that want generated video kept within their own systems. Outputs arrive as drafts that a team reviews, shapes and approves before publishing, which keeps the owner in charge of what goes out.

Alibaba: official site

Wan in plain language

Many current video models start from random visual noise and refine it step by step until the frames match a written description or a reference picture. The model learns those patterns from large collections of footage, so it can produce motion that looks convincing over a few seconds. Fine detail such as logos, hands and backgrounds is where a clip is refined, so each output is best treated as material to shape rather than a finished asset.

Two inputs shape most results. A written prompt sets the subject, setting, camera movement and mood, and its wording matters a great deal. A reference image fixes much of the look, so starting from a clean product photograph usually gives more control than describing everything in words. Clip length and resolution are also limited next to a filmed shoot, which means the output suits concepts, loops and background elements better than finished advertising.

Running the model yourself changes where the work happens. Someone must own the machine, its graphics memory and the software around the model, and keep those working as versions change. A hosted service takes on those duties, so the choice comes down to who carries the upkeep and where prompts and clips are stored.

What businesses use it for

  • Short clips showing a product in use, drafted from existing photos
  • Motion versions of listing images for property or catalogue pages
  • Rough video concepts to test before a full shoot is booked
  • Looping social media drafts that staff review before posting
  • Mood footage for newsletters, pitch decks and event slides

Who it suits

  • Retail and ecommerce teams that already photograph their products
  • Marketing teams testing video ideas before committing to a shoot
  • Owners who want generated footage to stay on systems they control

How MILabs sets it up

The easiest first step is to say hello through the enquiry form. The owner or team lead then decides which content the business wants to produce and how often. Together you choose where Wan will run, whether on a machine the business already owns or on a server provisioned for it. Next, MILabs connects Wan to the places the business already keeps its work, such as a shared drive for source photos, the file store for drafts and the social or web tools where approved clips go. Access is then set so only named staff can submit prompts and approved users can download outputs. Finally, you agree who looks after the installation, who applies updates and who reviews drafts before they leave the business.

Builds it fits

Industries it fits

Wan in practice

Inside a retail team, the work usually begins with a folder. Someone selects product photographs cleared for use, checks the rights to each one, and places them in the source folder the setup reads from. A staff member then writes a short prompt describing the movement wanted, such as a slow turn around a bag on a neutral surface, and submits it through the agreed interface. Output lands in a review folder before anything reaches a publishing queue.

A second person reviews each clip against a short checklist. They look at edges, on-screen text, colour against the real product and any feature the clip might imply that the product lacks. Approved clips move on to the social or product page tools, where a caption is written separately. Rejected clips are logged with the reason, so the next prompt improves rather than repeats.

For a marketing team testing concepts, the same routine works as a sketchbook. Several variants of one idea are generated, the group picks one or two, and only those reach a real shoot budget. Prompts sit beside their results so the team can see which wording produced usable motion. Across repeated use, that record becomes the team's own guide to where the tool helps.

good questions, straight answers

Frequently asked questions

How does MILabs set up Wan?
MILabs sets up Wan after an enquiry, beginning with the owner or team lead. Together you agree which content the business wants, where the model will run and who can submit prompts. The setup then connects Wan to the shared drives and publishing tools the business already uses, sets permissions for each staff role, and records who maintains the installation and reviews drafts. Scope is agreed in writing before any work begins.
Where does Wan store data?
Wan is released as open model weights, so a business running it on its own hardware or server keeps prompts, source images and generated clips in places it controls. MILabs records where those files sit during setup, including the folder structure and any backup location. The business decides retention periods, deletion steps and who can access archived outputs, and these go into the setup notes. A hosted run keeps files with that provider.
When is Wan the wrong choice?
Wan suits concepts, loops and background footage best. A business that needs a finished, edited film will want a person working in a video editor alongside it. Where a hosted tool with a clearer support route suits the business better, MILabs will say so during the enquiry and recommend that route. Teams with a machine ready for the model can host it, and the rest can use a managed service.
Can my team use Wan without training?
Yes, staff who only review drafts can usually work without any technical background. Staff who write prompts and manage the installation need a clear written routine and some hands-on time with real files. MILabs keeps the setup simple by limiting who can submit prompts and by documenting the steps each role follows. How the team learns the tool is a decision for the owner, who knows the team's time and skills best.
Can MILabs build on top of Wan?
Yes. MILabs can build workflows that call Wan as one step, for example taking an approved product photo, requesting a short clip, and placing the draft in a review folder for a named person. The scope of any such build is agreed after the enquiry and depends on the systems the business already uses. Each build logs what it sends to Wan and what it returns, and who signs off on it.
let's make something happen

Workflow automation: how we can help

Choose how you want to work on this with MILabs. Both options start with an enquiry.

Discovery is free, no obligation from the first chat.

Let us build it for you

We build it and look after it.

Build it yourself, we'll help

You build it with our direction, teaching and a weekly check-in.