Wan Image
Alibaba's Wan 2.7 picture model: consistent sets of pictures, long text in 12 languages and fine instruction following, at a low price per picture.
What it nails
- Consistent sets of the same character
- Text in 12 languages
- Follows long, detailed briefs
Watch out for
- No reference pictures in the Lab yet
Wan is Alibaba's creative family, known for video, and Wan 2.7 Image (announced 1 April 2026) is its picture model. Alibaba built it around three things: text rendering, subject consistency and complex instruction following. Ask for a set of pictures of the same character and it keeps the face; ask for long text and it renders it in 12 languages.
The Lab runs Wan 2.7 for text-to-picture, up to 4 pictures per run.
At a glance: Wan. The live spec card, price calculator and Generate button show here on the site.
What Wan 2.7 is secretly great at
Picture sets with the same character
Describe each picture in one prompt ("First image… Second image…") and ask for the features to stay consistent. Set pictures to 4 in the studio and you get the whole series.
A cinematic group of images documenting the same stray ginger cat. The features must be consistent throughout. First image: In spring, the ginger cat weaves through blooming cherry blossom trees. Second image: In summer, the ginger cat cools off in the shade on an old street. Third image: In autumn, the ginger cat steps on a ground covered with golden fallen leaves. Fourth image: In winter, the ginger cat walks on the snow, leaving footprints.
Example by Alibaba · source ↗
Pictures that explain a process
Wan follows long, detailed briefs: a left-to-right flow, a human for scale, empty space for your headline. That makes it strong for campaign and explainer visuals.
Premium landscape advertising key art for an industrial water-recycling campaign, shown as a cinematic architectural cutaway of a contemporary textile dye house. On the left, broad ribbons of fabric in saturated vermilion and deep indigo sweep through stainless-steel dyeing machinery; colored wastewater travels through transparent pipes into a sculptural central filtration system with stacked circular membranes, glass tanks and precise valves. Across the image, the liquid visibly transforms from dark dye into luminous clear aqua water, then loops back toward the production line on the right, making the closed-loop process instantly understandable. A female textile engineer in cream workwear stands beside the filtration unit for human scale, inspecting a tablet without dominating the composition. Art direction: sophisticated photorealistic CGI blended with crisp editorial infographics, immaculate industrial design, tactile fabric weave, condensation on glass, brushed metal, subtle steam, realistic reflections. Deliberate restricted palette: deep indigo as the dominant factory environment, warm cream architecture, vermilion fabric accents, bright aqua recycled water, and charcoal machinery. Strong lateral flow from left to right, rhythmic pipework and fabric curves, generous uncluttered cream wall area in the upper right for campaign copy, wide 3:2 composition, polished global sustainability advertising aesthetic, optimistic and technically credible, soft directional skylight with focused highlights, no logos, no labels, no typography, no UI, finished standalone campaign image.
Example by Runware · source ↗
Text inside the design
Alibaba says Wan 2.7 renders text in 12 languages. Put the words in quotes and say where they go and in what style.
Lookbook photo of a model wearing a [dark leather bomber jacket], minimalist studio, soft side light, confident pose, magazine cover aesthetic, clean backdrop, subtle film grain, high fashion. With a "Fashion Magazine" book name in art word style at the bottom
Example by WaveSpeedAI · source ↗
One idea, any style
Keep the subject, change one phrase ("in a photorealistic style", "in a pencil drawing style") and Wan switches the whole look.
An astronaut riding a horse in a photorealistic style.
Example by fal.ai · source ↗
An astronaut riding a horse in a pencil drawing style.
Example by fal.ai · source ↗
Wan 2.7 in the Lab
| Setting | What you get |
|---|---|
| Pictures per run | Up to 4, ideal for a consistent series |
| Shapes | Square, 4:3, 3:4, 16:9, 9:16 |
| Reference pictures | Not in the Lab yet: Wan runs text-to-picture here |
Wan 2.7 does not rewrite your prompt and does not take a negative prompt, so write exactly what you want to see, and phrase exclusions positively ("an empty street" instead of "no cars").
How to prompt Wan 2.7
From Alibaba's Model Studio guides.
- Say it all: subject, scene, style, lighting and composition.
- Write exactly what you want. There is no automatic prompt rewriting.
- For a set, number the pictures ("First image… Second image…") and ask for consistent features.
- Quote the text you want in the picture and say where it goes.
Beauty and fashion close-ups
Close-up portrait of a model whose face is partially covered in flowing liquid metal or an iridescent, second-skin-like substance. She has otherworldly, light purple eyes and stares directly into the camera. The background is completely blurred out, leaving only a soft halo of light. The lighting is even and ethereal, as if from a bioluminescent source. Inspired by the style of Nick Knight, the image emphasizes surreal textures and subtle color gradients, exceptionally sharp, with breathtaking detail, 16K.
Example by WaveSpeedAI · source ↗
FAQ
Is this the Wan video model?
No. This guide covers Wan 2.7 Image, the picture model. Wan video has its own guide.
Can Wan make a set of matching pictures?
Yes. Describe each picture in one prompt, ask for consistent features, and set the studio to 4 pictures.
Can I upload a reference picture to Wan?
Not in the Lab yet. For reference-based work, use Nano Banana, Seedream or Kling Image.
Which languages can Wan write?
Alibaba says Wan 2.7 renders text in 12 languages.