AI video with WAN — models, capabilities and examples

WAN (also written Wan) is a family of artificial intelligence models for generating video, developed by Alibaba through its Tongyi research lab (the project is also known as Tongyi Wanxiang). The models turn a text description or an already existing video into a new, moving clip, understanding the content of the scene, the motion and the style.

A distinctive trait of WAN is that the models are open source and are distributed under the free Apache 2.0 licence. That is precisely why WAN became one of the most widely used open video models in the world — a large community has grown around it, and developers and platforms all over the world integrate it into their own tools.

On genkiki.com the models of the WAN family are connected directly into the AI Studio, so you can use them through your browser, in your own language, with no complicated installation or a powerful computer of your own. Below are the specific variants we offer and what each of them is for.

Versions and options on generiram

On generiram there are three WAN variants available, covering two main working modes — creating video from text and transforming an existing video.

VariantModeWhat it does
WAN T2VText to videoCreates an entirely new video from your text description alone — you set the scene, the action and the mood with words, and the model turns them into footage.
Wan 2.2Video to videoTransforms a video you supply into a new version, following the motion and the composition of the original and giving them a different look or style.
Wan 2.6Video to videoA newer variant for working with video, which also starts from a clip you supply and creates a new version following your instructions.

Text to video (WAN T2V)

The “text to video” mode suits you best when you have no source material and want to create something from scratch. You describe what should be seen — characters, setting, motion, atmosphere — and the model generates the clip for you. It is a fast way to visualise an idea that you would otherwise have to film or draw.

Video to video (Wan 2.2 and Wan 2.6)

In the “video to video” mode you start from a clip you already have. The model keeps the motion and the structure of the footage but gives it a new look according to your instructions. Wan 2.6 is the newer variant — we offer both, so you can choose whichever result you like better for the task at hand.

Strengths

  • An open model with a wide community. WAN is among the most recognisable open video models, which means active development and technology proven in practice.
  • Natural motion. The models do well with smooth, consistent movement between the frames, so the scenes look whole rather than like a series of unconnected pictures.
  • Following the instructions. WAN tries to stick to the text description or to the video you supplied, which gives you more control over the final result.
  • Flexibility. With the two modes you cover both creating from scratch and reworking existing material — from one and the same family of models.

What it is good for

WAN is a handy choice in many everyday content situations:

  • Short clips for social media — dynamic footage for Instagram, TikTok, YouTube Shorts and Facebook.
  • Advertising and product visuals — turning an idea into a moving scene quickly, without a shooting day.
  • Creative and artistic projects — visualising concepts, moods and styles.
  • Reworking existing videos — giving a new look to a clip you already have, through the “video to video” mode.
  • Prototypes and drafts — a quick visual check of an idea before you invest in a full production.

How to start

To try WAN, go into the genkiki.com AI Studio and pick the variant you want — WAN T2V to create video from text, or Wan 2.2 / Wan 2.6 to transform an existing video. Then describe what you want (or upload your source clip) and start the generation. Everything happens in the browser, in your own language, and you can download and share the finished result. If you are not sure where to begin, try a short, clear description of a single scene and gradually add details until you hit the look you want.

A(z) WAN modelljei — részletesen

Melyik mire jó, miben gyenge, és mire nem érdemes használni.

Wan 2.2

videóvideó
Képarányok: 1:1, 9:16, 16:9 Hosszok: 5 mp, 10 mp Ár: 155 kredit

Erős ebben

  • It reworks finished video with precise technical control
  • TWO specialists sit underneath: one lays out the scene, the other perfects the texture and the detail
  • Trained on a curated aesthetic with labels for light, composition, contrast and colour — that is why it listens to instructions about the look
  • Strong on faces, hands and athletic movement
  • An open model with a free licence

Gyenge ebben

  • It wants more patience and a more precise description than the fast models
  • Its native clips are short
  • The resolution falls behind the 4K models

Ne használd erre

  • Don’t use it for a quick, rough change
  • Don’t use it for a long clip
  • Don’t write a scattered description — it punishes imprecision

Tipikus feladatok

  • A precise rework of a filmed clip
  • Changing the light and the mood of finished video
  • A shot with people where the face and the hands have to hold up

Wan 2.6

videóvideó
Képarányok: 1:1, 9:16, 16:9 Hosszok: 5 mp, 10 mp Ár: 194 kredit

Erős ebben

  • Reworking video from a REFERENCE — you show it how the result should look
  • It leaves the movement of the original untouched
  • It inherits the family’s strong aesthetic — light, composition, colour
  • Good with people: face, hair and hands stay convincing

Gyenge ebben

  • The result depends heavily on the quality of the reference
  • Slow
  • Short clip

Ne használd erre

  • Don’t feed it a blurry or tiny reference
  • Don’t use it for a quick change

Tipikus feladatok

  • Change the look of the character in a finished clip
  • Transfer a style from a reference onto filmed video

Próbáld ki most: WAN

Saját videó WAN modellel, egyenesen a böngészőben — telepítés nélkül, magyarul.

AI stúdió megnyitása Árak megtekintése