Adobe MAX: ChatGPT and Claude will soon be able to drive Photoshop

At its MAX conference in Miami, Adobe added a long list of new AI features to Photoshop, Premiere, Lightroom and Firefly. The biggest news is a local MCP interface that will let AI assistants from OpenAI, Google and Anthropic work directly inside Adobe project files. A private beta starts in early 2027.

Projects Nov 10, 2026 6 min read Source: Merkliste: hyper.ai
Illustrative image, AI-generated

What's happening

On November 10, 2026, at MAX, its annual creative conference in Miami, Adobe announced a long list of new Creative Cloud features. They cover Photoshop, Premiere, Illustrator, InDesign, Lightroom, After Effects and the Firefly AI platform. The theme is the same across the board. According to The Verge, Adobe is going all in on "agentic" AI: assistants that don't just generate a single image but handle multi-step tasks on their own. 9to5Mac reports that many of the features are rolling out now, while others are labeled beta or preview.

The announcement likely to matter most is an interface that lets third-party AI assistants such as ChatGPT, Claude or Gemini operate Adobe apps directly. You'll have to wait until 2027 for it, though.

MCP: outside AI agents working in Adobe files

Adobe is building local integrations for the Model Context Protocol (MCP). MCP is an open standard that gives AI assistants access to apps and data, a kind of universal plug between a chatbot and software. According to The Verge, external agents from OpenAI, Google and Anthropic will be able to run locally in Photoshop, Illustrator, Premiere and After Effects and edit native project files on your own machine.

Adobe stresses that everything stays editable:

"Because it runs locally on your computer, your agent works directly in your native project files, and your layers, vectors, bins and sequences stay editable."

That means you can step in and make manual changes at any point. Agents can also connect to other tools over MCP, so entire workflows become possible: pull reference documents from Google Docs, build a project from them in Adobe apps, then send the result to colleagues or clients via Slack. 9to5Mac gives the example of using ChatGPT to adapt a whole ad campaign to a new key visual you designed in Photoshop.

Adobe vice president Deepa Subramaniam told TechCrunch the company hasn't yet settled which functions it will expose:

"We're still figuring out where, when and how we draw the line. […] We have no need or desire to limit what the MCP endpoint can expose."

At launch there will be a set of endpoints for the four apps, and Adobe plans to add more based on usage. None of this is available yet. Adobe says the private beta begins in early 2027, and you can sign up in advance. TechCrunch notes that ChatGPT and Claude integrations already exist for some individual features.

New tools in Photoshop, Premiere and Lightroom

Photoshop gets Relight, a tool for adding, moving and resizing light sources after the fact without touching the original. Generative Restyle creates style variations of an image from a text prompt, which you can then fine-tune with sliders. There's also a before-and-after slider (Quick Compare). The beta app adds, among other things: Mockups for realistically placing designs on objects, Camera Angles for new perspectives on an image, Generative Recompose for reworking an image into a different aspect ratio, Convert to Text Layers (text in a flattened image becomes editable again), Timelapse and Remove Ambient Effect Only. That last one removes lens flares, haze or stray shadows without touching any objects.

Premiere gets tools for raw footage: Enhance Audio, Separate Crosstalk (which splits overlapping voices) and Separate Sound Source. Adobe says Generative Remove (beta) erases distractions such as boom mics, logos or background clutter right in the timeline. Color Mode is available now, a new color grading system that handles everything from quick fixes to elaborate looks. The Generative Media Tool brings video, sound effect, ambience and music generation directly into the timeline. That lets you generate missing B-roll (supplementary footage) to fit your edit, using Firefly or models from Google, Kling, Runway and Luma.

Premiere Mobile on iPhone now offers hundreds of thousands of licensed songs through partnerships with Universal Music and Warner Music, but only for non-commercial use. Coming soon: Quick Cuts, which builds a rough cut from your footage and a prompt, plus an AI assistant.

Lightroom gets Glow for soft, cinematic light, Blemish Removal for skin retouching, Open Eyes (beta) to fix closed eyes, and expanded tethering, meaning shooting with a camera connected directly to your device. That now works on the iPad too, with a live preview. Illustrator gets new brushes, gradients, patterns and blend controls. InDesign adds co-editing (beta), the adaptive Flex Layout, and Check Document and Rewrite for catching editorial and accessibility issues.

Topaz Labs, which Adobe recently acquired, contributes two image models. Wonder 4 handles upscaling, denoising, deblurring and sharpening in a single step. Image HDR converts standard images to HDR, meaning images with a wider brightness range.

Firefly becomes an agent-powered workspace

Adobe wants Firefly to go from a standalone AI generator to a full creative environment. Firefly Video arranges scenes, assembles clips, adds captions and generates transitions and B-roll. Firefly Design puts illustration, image editing, animation and 3D on a single canvas. Also new: Quick Actions for common jobs like removing backgrounds or creating video thumbnails, Firefly Graph for reusable workflows, and Firefly Builder (private beta). Builder lets you prototype websites and apps in plain language. TechCrunch sees this as Adobe moving onto rival Figma's turf.

The Firefly AI Assistant (beta) gets a Plan Mode, where you review, edit and approve the agent's plan before it starts. With the Model Selector, you pick the AI model yourself or let the system choose. You can also store brand guidelines, colors, characters and products and have them applied automatically to every project. As a preview, Adobe showed always-on agents that handle recurring tasks continuously in the background. The Verge's example is an agent that automatically turns every incoming image into ad graphics for different social networks, leaving you to pick the best results each day. There's no release date yet. The feature will come to the Firefly AI Assistant first.

Adobe's "Creative Agent" is also available as a beta assistant in Illustrator, Premiere, InDesign, After Effects, Frame.io and the Photoshop beta. In Premiere, for example, it can track masked subjects across shots or adjust color and exposure across entire sequences. For enterprises, the Economic Times reports that Adobe announced the "CX Enterprise Coworker" and "Brand Intelligence AI" for consistent branding as part of GenStudio, among other things.

The big picture

Adobe is sticking firmly to the course it set in April. Back then, creative chief David Wadhwani said the Creative Agent would show up "in virtually everything we build." Opening up to outside agents via MCP is a notable strategic move. Instead of locking users into its own assistant, Adobe wants to make sure its apps stay essential in workflows built around ChatGPT, Claude or Gemini. Shortly before the event, Figma co-founder Dylan Field argued that MCP isn't a perfect standard and mostly benefits the big AI labs.

Many of the new features are still in beta, in preview or slated for 2027. It remains to be seen how reliably agents perform in complex professional projects and how much control you actually keep over automated changes, a point the Economic Times also makes. It's also unclear which functions the MCP interface will actually expose at launch. Adobe says it hasn't decided yet.

Sources

#adobe#mcp#ai agents#photoshop#firefly#premiere
Read the original source

Ask the AI assistant

Questions about this article, installation or background? Our assistant knows this page and runs on our own machine, not in someone else's cloud.

Connecting …

The Assistant is an AI (Gemma 4) and can make mistakes – please check important details against the source. Questions are used only to generate an answer and are deleted after one hour. Please don't enter any personal data.

Comments

What do you think? Already tried it? Write it here, no sign-up needed.

  1. Loading comments …

Comments are public. Links aren't allowed, and we delete abusive posts. You can delete your own comment at any time in this browser. Learn more in the Privacy Policy.