Pro-funktionWan Image is Alibaba’s Tongyi Wanxiang image model—describe a finished frame from text, or upload a reference photo for Image to Image edits. Wan 2.7 adds Thinking Mode for composition planning, twelve-language in-frame type, HEX palette control, and multi-reference fusion; Wan 2.5 and 2.6 tiers stay useful when you need faster drafts.
12
DemoSee a Wan 2.2 text-to-image training walkthrough—how the model handles image generation—then draft your own Wan Image prompt with Thinking notes in the tool above.
The samples below cover product heroes, atmospheric street scenes, character consistency, e-commerce flat lays, golden-hour landscapes, and bilingual signage—each with a prompt you can adapt in the tool above.
Clean product isolation starts with surface, prop layout, and negative space—name each before you lock a catalog frame.
Matte black wireless earbuds with charging case on warm linen, soft daylight, photoreal e-commerce catalog hero, generous padding, 16:9.
Atmospheric street scenes need layered lighting, wet pavement, and vendor action—describe each before you pick resolution.
Outdoor night market alley at dusk, square crop, photoreal cinematic wide shot, warm string-light glow across BBQ stalls, vendors prepping skewers, atmospheric steam, wet stone pavement, deep teal sky.
Multi-reference fusion keeps one character readable across scenes—list wardrobe anchors and what must not drift per scene.
Young professional woman, same outfit and hairstyle across three scenes: café window, rooftop walk, evening bookshelf—photoreal lifestyle trilogy, consistent face, 16:9 each.
Image-set mode rewards one lighting recipe across SKU tiles—brief props, surface, and HEX palette once.
Overhead e-commerce flat lay: serum bottle, dropper, and botanical sprig on cool marble, consistent soft daylight, square tile, full products visible with padding.
Atmospheric landscapes reward Thinking Mode—lock horizon line, light direction, and color palette before you render wide outdoor scenes.
Rolling lavender field at golden hour, soft purple rows, distant tree line, gentle hills, cinematic depth, natural colors, 16:9 landscape photograph.
Bilingual storefront type needs both strings, font weight, and safe margins above the door line.
Vertical boutique storefront at dusk, neon accent light, readable bilingual sign: English OPEN and Japanese 営業中 beneath, clean sans-serif, safe margin above door, photoreal street scene.
Wan Image brings thinking-first composition, twelve-language type, HEX palette control, multi-reference fusion, image sets, and 4K Pro exports. Below maps each capability to brief fields VidMage tracks before you generate.
Wan 2.7 Image includes Thinking Mode: the model plans composition, lighting, and subject placement before rendering—note thinking on/off in VidMage when speed beats reasoning depth.
Wan 2.7 Image accepts up to nine input images for edit or fusion—brief one scoped change per pass and list what must not drift in faces, product geometry, or wardrobe.
Wan-Image and 2.7 pages stress ultra-long in-frame copy—provide verbatim headline, label, and footnote strings with scale and safe margins when legal or brand review depends on readable details.
Wan 2.7 Image supports HEX palette control—list primary, accent, and background hex values in the brief so brand teams do not rely on vague color adjectives alone.
Wan 2.7 Image supports image set mode —one prompt can return several related images when you label each frame (first image: …, second image: …) so SKU rows, colorways, or storyboard beats share the same lighting and product geometry.
Stack product, material, and palette references—then fuse them into one shelf-ready catalog hero. Pick 4K Pro when print or marketplace specs need extra pixels; 2K covers most social and web crops.
VidMage is where you lock reference-stack roles, HEX palette lines, thinking on/off intent, twelve-language type strings, version notes (2.5 vs 2.7), and moderation guardrails—then hand a review-ready brief to wan.video or your API runner instead of improvising across separate tabs.
One upload, one change list—add prop, swap background, relight—so Wan Image edits do not drift into full-scene regenerations.
Poster headlines, SKU labels, and footnotes live as copy-paste strings—not paraphrased placeholders—for brand and legal review.
List primary, accent, and background hex values so Wan Image hosts do not guess brand colors from mood words alone.
Number each panel in a set-mode brief so batch output does not shuffle props or lighting mid-grid.
Note whether the job runs on wan.video, Replicate, or your API runner—and stick to official apps for billing or policy questions.
Mirror likeness, celebrity, and regulated-vertical notes from host Terms into the brief before generation—not after a block.
Export prompt, references, thinking notes, and chosen variants as a review bundle instead of scattered host tabs.
When Wan Image exports feed the WAN AI video page or Seedance tools, keep crop, type, and character notes attached to the exported frame.
Compare Wan 2.5 Image, Wan 2.6 Image, and Wan 2.7 Image on reference inputs, resolution, thinking mode, typography, fusion limits, and edit style—so you can pick the right version for your brief.
| Features | Wan 2.5 Image | Wan 2.6 Image | Wan 2.7 Image |
|---|---|---|---|
| Reference inputs | Text and single image | Text and multi-image refs | Text and up to nine refs |
| Max resolution | Up to 2K | Up to 2K | 2K standard; 4K Pro |
| Thinking mode | — | Limited | Yes — plans before render |
| Text & typography | Good short type | Better multilingual lines | Twelve-language long copy |
| Reference fusion | Fewer upload slots | Multi-ref edit | Up to nine refs |
| Native alpha | — | Partial | Yes |
| Edit & extend | Scoped single-upload edit | Multi-image fusion | Instruction edit + set mode |
| Best for | Rapid drafts and mood boards | Campaign images with refs | Posters, product sets, precise type |
| When to pick | You need fast exploration | You have a reference stack | You need thinking, type, or 4K |
Use VidMage to prepare your Wan Image brief: pick Text to Image or Image to Image, write the finished frame, note Thinking or aspect-ratio needs, generate, then refine the next direction.
Start with Text to Image when the idea can be described from scratch. Use Image to Image when a product photo, moodboard, style frame, or previous output should guide the result.
Describe the subject, medium, setting, lighting, color mood, composition, style reference, aspect ratio, and details that should stay recognizable.
Preview the image set, compare variations, tighten the prompt, adjust crop or reference notes, download the strongest frame, and refine the next direction.
Wan Image fits e-commerce leads, brand designers, and studio teams shipping catalog heroes, poster comps, lifestyle campaigns, and trend-reactive social frames—with thinking-first composition and multi-reference fusion on Tongyi hosts.
Draft product visuals, ad concepts, seasonal campaign images, social posts, and landing-page art before final design polish.
Explore art direction, moodboards, key visuals, poster styles, and concept frames when texture, lighting, color, and taste carry the idea.
Turn quick ideas into portrait, square, and vertical image concepts for Instagram, Pinterest, thumbnails, and hook testing.
Compare alternate hero frames, campaign routes, pitch-board visuals, and platform-specific image directions before production starts.
Shape visualization concepts, room moods, material studies, and client-presentation images from text plus reference descriptions.
Explore cover art, character studies, environments, color systems, and moodboards, then use the strongest direction as a production reference.
Type a new scene or upload a photo to edit. Add lighting, crop, or on-image text, then hit Generate when you're ready.
