Wan 2.7 · 生成も編集も 1 モデルで
アリババの通義万相 Wan 2.7 が OVOV に登場。1 つのモデルで文生画像と画像編集の両方に対応し、中国語・バイリンガルの組版に優れ、参照画像 9 枚まで、鮮明な 2K 出力。
各モデル名およびロゴは、それぞれの所有者の商標です。OVOV は独立した製品であり、いかなるモデル提供元とも提携・推奨・スポンサー関係はありません。
Why Wan 2.7 is different
Built by Alibaba to be a versatile all-rounder — one model that generates from text and edits existing images, with typography that finally gets Chinese right.
Generate and Edit in One Model
Text-to-image and image-to-image live in the same model. Start from a prompt or restyle an existing photo — no switching tools.
Chinese and Bilingual Text
Native-quality Chinese typography plus clean English — ideal for posters, packaging and social graphics for Chinese-speaking audiences.
Up to 9 Reference Images
Feed up to 9 references to keep a character, product or style consistent, or to guide precise edits.
Strong on Oriental Aesthetics
Ink-wash, guochao, festival and traditional motifs render with a cultural fluency Western-trained models often miss.
Crisp 2K Output
Sharp 1K and 2K exports across common ratios — square, landscape, portrait — ready for web, print and social.
Reliable Prompt Following
Wan 2.7 stays faithful to your composition, subject count and layout instructions — fewer surprises, more usable first drafts.
From prompt or photo to finished image
Wan 2.7 handles both paths — generate fresh or edit what you have.
Describe or upload
Write a prompt in English or Chinese, or upload an image (plus up to 9 references) to edit.
Wan 2.7 renders
The model composes your scene or applies your edit, rendering in 1K or 2K.
Download, no watermark
Export commercial-grade images with full usage rights — no watermark, no cleanup.
What you can build
Wan 2.7 is the versatile pick for bilingual content, cultural aesthetics and everyday generate-or-edit work.
Chinese & Bilingual Posters
Festival posters, promotions and packaging with flawless Chinese set alongside English.
- Native-quality Chinese typography
- Bilingual layouts in one prompt
- Guochao and festival motifs
Product & Social Graphics
E-commerce visuals, banners and social cards across square, landscape and portrait ratios.
- Common ratios for every channel
- Crisp 2K for web and print
- Faithful subject and layout control
Photo Editing & Restyling
Restyle, extend or refine an existing image using up to 9 reference images for guidance.
- Text-to-image and editing in one model
- Up to 9 reference images
- Precise, controllable edits
Oriental & Artistic Styles
Ink-wash landscapes, traditional illustration and cultural artwork with authentic detail.
- Ink-wash and traditional motifs
- Cultural fluency out of the box
- Artistic and editorial ready
Prompt writing guide for Wan 2.7
Wan 2.7 reads natural language in English and Chinese — these techniques get the most from its text rendering and editing.
Best with clear subject and style cues
Wrap exact text in quotes — English or Chinese
Quoted strings are rendered literally. Wan 2.7 is especially strong with Chinese characters.
- •Text inside quotation marks is rendered as-is
- •Mix Chinese and English in one prompt
- •Combine multiple quoted strings for multi-line copy
Example
"A festival poster reading "新年快乐 · Happy New Year" in bold red and gold"
Upload references to edit or keep consistent
Provide up to 9 reference images to edit an existing image or lock a subject across a series.
- •Use clean, well-lit references
- •Describe what to change vs. keep
- •Great for restyling and product consistency
Example
"Restyle the uploaded photo as a traditional ink-wash painting, keep the composition"
Name the artistic style explicitly
Wan 2.7 excels at Oriental and traditional styles — call them out for authentic results.
- •Name styles like "ink-wash", "guochao", "ukiyo-e"
- •Add era or regional cues for accuracy
- •Combine style with a clear subject
Example
"A guochao-style illustration of a phoenix, red and gold palette, intricate linework"
Specify subject count and layout
State how many subjects and where they sit. Wan 2.7 follows composition instructions reliably.
- •Say "two people", "centered", "rule of thirds"
- •Describe foreground vs. background
- •Name the aspect ratio you want
Example
"Two figures centered, mountains in the background, 16:9 landscape"
Ask for 2K for final output
Request 2K for finished web and print work; 1K keeps drafts fast while you iterate.
- •State "2K" for final, sharper output
- •Use 1K while exploring concepts
- •Richer descriptions reward higher resolution
Example
"2K e-commerce hero of a teapot on a wooden table, soft daylight"
State constraints at the end
Tell the model what to leave out — extra text, watermarks, clutter. Direct negatives work best.
- •Append "Constraints:" at the end of the prompt
- •Forbid extra text, watermarks and logos explicitly
- •Avoid vague phrasing like "try not to"
Example
"Constraints: no extra words, no watermark, clean uncluttered background"