Adobe brings an AI assistant to Photoshop

Editing in Photoshop used to mean wrestling with complex adjustment layers and pen tools. Adobe wants to change that dynamic by letting users just describe what they want or point directly at the canvas.

Adobe rolled out a public beta of AI Assistant in Photoshop across web and mobile platforms. The company also pushed a major upgrade to its standalone Firefly Image Editor, bringing unified editing tools and direct access to more than 25 external generative AI models.

To drive adoption, Adobe is offering paid Photoshop subscribers unlimited generations with AI Assistant through April 9. Free-tier users get 20 complimentary generations to test the waters.

Talk to your canvas and mark it up

The AI Assistant essentially turns Photoshop into a conversational partner on web, iOS, and Android.

On mobile devices, you can simply speak your edit requests out loud. Tell the app to clear out background tourists, enhance shadows, or crop the photo like an indie album cover. The assistant can execute the changes automatically or walk you through each step if you want to learn the actual workflow.

Over on Photoshop web, Adobe added a new feature called AI Markup in the contextual task bar. Instead of fiddling with delicate selections, you sketch directly onto any area of the image and type a prompt, such as adding a patch of wildflowers or replacing a background building. The engine fills in the marked zone within seconds.

Firefly turns into a multi-model editing hub

At the same time, the Firefly Image Editor received a major tool consolidation.

Features that were previously fragmented are now housed under one streamlined roof, including Generative Fill, Generative Remove, Generative Expand, Generative Upscale, and one-click background removal.

Even more notable is Firefly's model roster. Creators can now toggle between more than 25 prominent models, spanning Adobe's commercially safe models, Google's Nano Banana 2, OpenAI's Image Generation, Runway's Gen-4.5, and Black Forest Labs' Flux.2 Pro. Users can generate an asset using an external model and transition straight into precision editing without leaving the workspace. Firefly subscribers also continue to enjoy unlimited generations, a perk introduced back in February.

Why it matters

We think this is one of the smartest strategic shifts Adobe has made in the generative era.

By pushing conversational editing and AI Markup straight to browser tabs and smartphones, Adobe effectively lowers the intimidating skill floor of professional editing. Students, marketers, and casual creators can now achieve production-ready visual adjustments without ever touching a clone stamp or a curve layer.

More importantly, Adobe is leaning into its role as an agnostic creative operating system. Rather than stubbornly walling users into its proprietary Firefly models, it is embracing heavy hitters like Flux, Runway, and OpenAI. Adobe knows that while anyone can train a decent image generator today, nobody matches its decades of pixel-level control and manipulation tooling.

By becoming the orchestrator where those external models get refined, Adobe protects its moat. The real test will arrive after April 9, when the promotional unlimited access expires and Adobe has to balance heavy third-party inference bills against creator subscription tolerance.