Adobe MAX 2026: AI Agents for Photoshop, Premiere and Firefly

Adobe put AI agents in Photoshop, Premiere and Firefly at MAX 2026 and deepened its Google deal. Here is what you can try now and what is still in beta.

5 min read

Adobe MAX 2026: AI Agents for Photoshop, Premiere and Firefly

Illustration made on Lumeta with GPT Image 2.5Make one like it

Adobe used its MAX conference in Miami on November 10, 2026 to put AI agents at the center of Creative Cloud. Its Creative Agent, an assistant you direct in plain English, is spreading into Photoshop, Premiere, Illustrator, InDesign, After Effects and Frame.io, and Firefly is being rebuilt as an all-in-one AI studio. On the same day Adobe and Google expanded their partnership, bringing Google's newest image and video models deeper into Adobe's apps.

For creators the headline is less a single new model and more a new way of working: you describe the job, and the agent does the clicks. The catch is that much of it is still labeled beta or preview, so it pays to know what is real today and what is a promise.

What changed in Photoshop, Premiere and Lightroom

  • Creative Agent in the desktop apps (beta): it carries out multi-step tasks from a written request. In Premiere it can apply an effect to every clip in a sequence, track a masked object frame by frame, and even out color and exposure across a whole edit.
  • Photoshop: Relight changes the lighting of a finished image, and Generative Restyle turns a prompt-based style into sliders you can fine-tune. Adobe showed both as coming soon rather than shipping.
  • Premiere: Color Mode, its grading system for professionals, took a turn on stage, alongside Generative Remove, which erases an object from video (in beta, for clips up to 10 seconds at up to 1080p, processed in the cloud).
  • Lightroom: Open Eyes, a fix for blinks in group shots, is in beta.
  • Sharper images: two models from Topaz Labs, which Adobe now owns, join the lineup. Wonder 4 upscales and cleans up noise, and Image HDR turns ordinary photos into HDR. Adobe has not given a date for them.
Paper-craft robot in a paint-splattered apron rolling paint on a framed mountain lake picture under a desk lamp, film strip nearby
A tiny helper touches up a mountain lake while the desk lamp does the relighting.

Illustration made on Lumeta with GPT Image 2.5Make one like it

Firefly becomes a studio

Firefly now gathers video assembly, image editing and design work in one place. Edit Object changes one part of a picture and leaves the rest alone. Quick Actions are one-step jobs such as removing an object, making a thumbnail or building a product ad. Firefly Graph (beta) lets you save a repeatable workflow, and Firefly Builder (private beta) turns a written idea into a clickable prototype of a website, deck or app. Adobe is also previewing always-on agents that tidy assets, check quality and prepare versions for each channel in the background.

Firefly keeps its model picker, with models from Google, OpenAI, ByteDance, Kling, ElevenLabs, Black Forest Labs, Runway and Topaz alongside Adobe's own. Partner models can carry their own usage terms, so check them before client work.

The Google deal, in plain terms

Adobe says customers can already use Google's Gemini Omni, Veo and Nano Banana inside its apps, and that Google's latest image and video models will reach Firefly, Photoshop, Premiere and Adobe Express. Acrobat is coming to Google Drive and Gmail as an "Edit with" option in file previews, Adobe's creative tools are coming to AI Mode in Google Search, and its connectors in Gemini will gain Acrobat and more creative tools. Neither company gave dates for these.

The two also said Google DeepMind and Adobe are exploring a custom Google image model for Photoshop, aimed at edits that happen in close to real time as you type. That is research, not a product announcement.

What it costs

Adobe has not announced prices for the new features. Generative tools draw on Adobe's generative credits: an Adobe team member said on its community forum that Premiere's Generative Remove uses 40 credits per second of video, so a 10-second clip costs 400 credits. Other credit costs were not published.

What you can do now

If you pay for Creative Cloud, open the beta apps and try the Creative Agent on a real but low-stakes project: it is best at repetitive jobs across many clips or layers. If you mostly want the Google models Adobe is leaning on, you can use them directly without a Creative Cloud plan: Nano Banana 2 for images and edits, Gemini Omni and Veo 3.1 for video, all on Lumeta.

Nano Banana 2Try the Google image model Adobe is building on: generate a picture, then edit one part of it with a sentence.

When was Adobe MAX 2026?

Adobe held MAX 2026 in Miami, with the main announcements on November 10, 2026.

Are the new Adobe AI agents available now?

Partly. The Creative Agent assistants are in beta in several Creative Cloud apps, always-on agents are a preview, and Firefly Builder is in private beta. Some features, such as Photoshop Relight, were shown as coming later.

Which Google models can I use in Adobe apps?

Adobe says Gemini Omni, Veo and Nano Banana are already available to its customers, and that Google's newest image and video models will come to Firefly, Photoshop, Premiere and Adobe Express.

Can ChatGPT or Claude edit my Photoshop files?

Not yet for most people. Adobe is previewing a local MCP connection for that, with a closed beta planned for early 2027.

Every top AI model, in one place

Nano Banana 2, GPT Image 2.5, Seedance, Veo and more, all on one credit balance. Start free with 35 credits a month. No card needed.

All tools for "From a description"

More from Lumeta

All News