Microsoft announces MAI Image 2.7 editing preview, with text-to-image still to come

Microsoft has announced MAI Image 2.7 as an image‑editing model in preview. Its October 27 post describes requests that change an existing asset: restyling or relighting a scene, changing text or symbols within an image, and restoring or enhancing material.

The examples also include moving a product into a different setting and reworking individual objects while accounting for the rest of the scene. Those are editing tasks, where the starting asset and the intended change both matter. For a creator, a useful brief identifies which elements should change and which should remain recognizable, then checks the resulting image against that brief. Microsoft’s announcement describes these use cases; OpenTools has not performed an editing test or verified a particular output.

Microsoft’s announcement names Arena, Artificial Analysis and MAI Playground as places to try the editing capability. Its actual Playground link, however, contains the parameter model=mai‑image‑2‑6.

In a public, read‑only inspection of that linked page, the model chooser and current image‑model label showed MAI‑Image‑2.6, alongside a separate promotion reading “Edit images with MAI‑Image‑2.7.” That is evidence of a public 2.7 editing entry being promoted, but it does not establish that the older link selects or runs 2.7. No account was used, image uploaded, model switched or generation submitted in this check.

Someone following the launch link should therefore distinguish the model named in the selected interface from the newer editing promotion. The issuer’s preview announcement and a working entry for every account are separate observations; the retained materials do not establish a universal regional or account‑access matrix.

The post explicitly says Microsoft is starting with editing, with text‑to‑image generation coming soon. It does not give a text‑to‑image launch date or announce a production API, pricing or usage limits for MAI Image 2.7.

Existing text‑generation controls or examples on a page labelled with another image‑model version should not be counted as proof that 2.7’s text‑to‑image mode has launched. The supported current story is Microsoft’s editing preview and its described creative tasks, with a public Playground promotion whose selected‑model path still needs to be distinguished from the announcement.

Microsoft Visitor Center Building 92, photographed May 29, 2022. Archival Microsoft company context, not an MAI Image 2.7 output, interface or demonstration. Photo: Lectrician1, via Wikimedia Commons; CC BY‑SA 4.0 International. Resized and cropped for display; image adaptations retain the same licence.