Wan 2.7 Pro Edit

• - 🏢 Alibaba's image editing and inpainting model, released April 2026.
• - 🔧 Handles instruction-based edits, local inpainting, outpainting, watermark removal.
• - 🖼️ Editing-focused sibling of the Wan 2.7 Pro image generator.
• - 📏 Outputs up to 2K resolution via Model Studio.
• - 🌐 Accepts multiple reference images for coherent composition.
• - 🎯 Pro tier targets high editing precision and multi-image consistency.
• - 💬 Edits driven by natural-language prompts, with optional masks.
• - 🔧 Available through Alibaba Cloud's DashScope SDK and HTTP API.

Alibaba

Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research,…

Wan 2.7 Pro Edit is the image-editing and inpainting member of Alibaba's Wan 2.7 image family, distinct from text-to-video and image-to-video siblings of the same generation. According to Alibaba Cloud's Model Studio documentation, the underlying wan2.7-image-pro endpoint supports text-to-image, image-to-image sets, generation from multiple reference images, and instruction-based editing, with output up to 2K resolution. It is positioned as the higher-precision counterpart to the standard tier, recommended for scenarios needing high editing accuracy or multiple coherent images.

As an inpaint-type model, its editing repertoire follows the Wan editing lineage: instruction-based edits without specifying a region, local inpainting within a masked area, outpainting/expansion, style transfer, watermark removal, and image restoration. Edits are driven by natural-language prompts and optional mask images supplied through the DashScope SDK or HTTP API.

Compared with earlier Wan editing releases — the Wan 2.1 general image-editing model documented by Alibaba offered inpainting, outpainting, watermark removal, and style transfer — the 2.7 generation extends the same instruction-driven workflow and adds multi-reference image conditioning at 2K output. It also pairs naturally with its generation-mate Wan 2.7 Pro for a generate-then-edit pipeline.

Within Alibaba's broader catalog, Wan 2.7 Pro Edit sits alongside other editing models such as Qwen Image 2 Pro and Qwen Edit 2511. I could not find official Wan 2.7 Pro Edit benchmark figures from Alibaba or a top independent evaluator, so this description stays limited to documented features rather than performance claims.