Adobe has significantly expanded its artificial intelligence capabilities across its core creative products, including Photoshop, Premiere, Illustrator, and InDesign, during its MAX event in Miami. The update introduces a new creative AI agent for Firefly and a Model Context Protocol (MCP) that enables users to leverage Adobe’s tools within external platforms such as ChatGPT, Claude, and Gemini.
The Firefly app is also venturing into prototyping with the launch of Builder, a tool currently in private beta that allows users to create designs to showcase concepts. Additional features include a tool for generating on-brand graphs and an image editor that can modify specific elements without affecting surrounding objects. Adobe is also debuting an AI-powered video assembler in Firefly, which compiles clips, images, and sound effects into finished pieces.
Specific updates to the Creative Cloud suite include new relighting features and prompt-based style sliders for Photoshop, a professional color grading system and object removal tool for Premiere, and enhanced capabilities in Lightroom such as blinking eye correction and blemish removal. The company is making its AI agent proactive and always-on to better understand user workflows, while also simplifying quick actions like background removal and thumbnail generation.
Adobe vice president Deepa Subramaniam noted that the company is still determining which specific endpoints and capabilities to expose via the new MCP across Illustrator, Photoshop, After Effects, and Premiere. The goal is to support creative workflows without imposing strict limits, relying on user feedback to refine the toolset. This move comes shortly after Figma co-founder Dylan Field criticized the MCP as an imperfect standard that primarily benefits frontier AI labs.
The update also includes new image models from Topaz Labs, a recent acquisition by Adobe. These include Wonder 4 for upscaling and denoising, and ImageHDR for converting standard images to HDR format, adding to the company’s existing integrations with models from Google, OpenAI, and other partners.
The blinking eye correction in Lightroom is actually useful. I was skeptical that they’d automate such a specific task.
Adding Topaz models for upscaling is a smart move. Quality has definitely improved in recent years.
Dylan Field’s criticism is valid. Standardizing MCP before it’s fully mature feels risky for the whole industry.
Love the specific image editor that modifies elements without affecting the background. That feature alone is worth the update.
Skeptical about these new agents. Hope they don’t start making creative decisions for me without asking permission.
Is the Model Context Protocol going to work with local LLMs, or just the big cloud models?
The video assembler in Firefly sounds like a huge time saver for my quick social media content needs.