Adobe Adds a Second Working Mode to Photoshop, Where Images Are Edited With Text

Adobe has released Photoshop 27.10, the August release of the editor that the entertainment industry uses for concept art, texture fixes, and render retouching. The main change affects the very construction of the program: Photoshop now has an alternative working mode, AI Assisted Editor, where the image is edited with text commands.

The manual interface remains under the name Pro Editor — the standard mode of manual editing. You can switch between the two ways of working right inside the app.

How the new mode is built and where its limits are

AI Assisted Editor is billed as a simplified editing mode with AI: edits are performed with natural-language commands. Judging by the description, it is aimed above all at Photoshop's existing generative tools — Generative Fill, Generative Expand, Generative Upscale, and Remove Background.

The work is declared non-destructive, with a caveat that matters to anyone handling complex multi-layer files. When moving from Pro Editor into the simplified mode, the document is flattened into a single layer; on returning, the result of the edit is added as a separate layer. Layer structure cannot be preserved inside the AI mode.

Formally, Adobe calls AI Assisted Editor a beta feature, although, judging by the release, it is available in the stable build rather than in separate beta versions for subscribers. The company has not offered any public clarification on this so far.

Edits by description have reached the regular interface too

Pro Editor has gained a Prompt to edit item in the contextual task bar. The logic is the same: the user describes what needs to be done, and Photoshop picks the appropriate tool itself. Besides obviously generative tasks such as changing a character's pose or adding rain and puddles, you can also ask it to change brightness or contrast.

Masks and markup over the frame instead of long prompts

A separate block of changes is devoted to the precision of generative edits. The area a generation applies to can now be set with a mask instead of a verbal description; the feature works when Adobe's own model — Firefly Image Model 5 — is used.

The second way is visual markup, as in software for review and collaborative work on a frame. The user draws arrows, circles, and “rough scribbles” over the image, specifying, for example, exactly where the generated object should appear, or leaves text notes. This capability is available when working with Google Gemini models, connected in Adobe products as partner models.

Part of the functionality is thus tied to Adobe's models and part to partner ones: users will have to keep track of which tool runs under which model.

Updates that are useful even without generative features

The release brought something for those who never touch AI as well:

• Light Adjustment Layer — a new adjustment layer for light: exposure, contrast, highlights, shadows, whites, and blacks. The edits are non-destructive and follow the standard logic of adjustment layers.
• Dynamic Text has learned to run text along an arbitrary shape or path — previously only circular and semicircular trajectories were supported.
• The list of formats available in the Save As command is now configurable.
• Adobe Stock content can be browsed and placed directly in Photoshop.
• Performance has been improved on processors of the Zen 4 and Zen 5 architectures — that is, on AMD chips less than four years old. The release notes do not state the size of the gain.

What it costs and what it will run on

Photoshop 27.10 runs on Windows 10 and newer, as well as macOS 14.0 and newer; in the documentation the version is listed as the August 2026 release. The distribution model is unchanged — subscription only. Photography plans with access to Photoshop cost $29.99 per month or $239.88 per year; a separate subscription to the editor is $34.49 per month or $263.88 per year. The full list of changes is published in Adobe's release notes.

Why Adobe needed a second interface

The arrival of a separate simplified mode looks like a response to changed user habits: a noticeable share of the audience came to image processing through generative services, where the only interface is a prompt line. For such a user, classic Photoshop with its panels and layers remains a barrier, and Adobe, judging by the release, is adding a second entrance to the same product instead of rebuilding the main editor.

The bet on masks and drawn markup also shows the limits of the approach: purely textual control gives too coarse a grip for professional work, so the familiar visual ways of indicating what to change and where are being brought back on top of the prompt. The prompt will likely settle in as an additional layer of control above the usual tools — production still needs layers, masks, and a predictable result.

Connecting partner models alongside its own points to a shift in the role of the editor itself: Photoshop is gradually becoming a shell through which various generative engines are accessible. For artists and studios that means a new item in the job requirements — understanding which model is responsible for a particular edit and what its limitations are.

Why did Adobe split Photoshop into two modes: pressure from generative editors, an influx of newcomers, a bet on subscriptions?