Key takeaways
- Adobe released an AI expansion across Photoshop, Premiere, Illustrator, and InDesign with Firefly as a proactive agent maintaining project context.
- The company launched a Model Context Protocol enabling ChatGPT, Claude, and Gemini users to access Creative Cloud features directly in those platforms.
- New capabilities include Photoshop's relighting and generative styling, Premiere's object removal and color grading, Lightroom's eye correction and blemish removal, and Firefly's video assembly tool.
At its annual MAX conference in Miami, Adobe announced sweeping additions of generative AI features to Photoshop, Premiere, Illustrator, and InDesign. Rather than confining AI to separate applications or isolated features, the company is embedding generative capabilities throughout its core creative tools and expanding its Firefly AI agent to work proactively within users’ existing workflows.
The strategy signals Adobe’s vision for AI in creative work: instead of replacing applications, AI should amplify creative decision-making at every stage of a project, from initial concept through final delivery.
Firefly Becomes a Proactive Creative Partner
Builder Enters Figma’s Design Prototyping Market
Adobe’s Firefly launched a feature called Builder, currently in private beta, that generates interactive prototypes from text descriptions. According to TechCrunch’s preview, Builder allows designers to create functional mockups to test ideas and gather stakeholder feedback without manually building prototypes in dedicated design tools. The feature directly competes with Figma’s prototyping capabilities while keeping workflow and assets within Adobe’s ecosystem.
Specialized Tools for Graphs and Selective Editing
Beyond prototyping, Firefly added a graph generation tool that creates visualizations matching a brand’s design language, reducing the need to export charts from spreadsheet applications or create them manually in Illustrator. A new image editing capability allows users to modify specific regions of a photograph without accidentally affecting surrounding areas, handling a common editing task that previously required careful layer masking or clone tool work.
Video Creation and Quick-Action Features
Firefly now includes an AI-powered video assembly tool that composes final videos from component footage, still images, and audio effects. This lets creators build videos without traditional video editing expertise or without opening a dedicated video production application. The company also prioritized making frequent tasks faster by automating background removal and thumbnail generation, operations that previously consumed time even though they involved limited creative decision-making.
More broadly, Adobe is making Firefly’s agent proactive and always on, meaning it maintains awareness of a user’s active project and can suggest relevant actions without requiring explicit requests.
Specific Enhancements to Individual Applications
Photoshop received relighting controls that adjust how light falls across a photograph after capture, along with sliders for applying AI-generated visual styles based on text prompts. These additions give photographers and designers control over lighting and aesthetic without leaving the application to process images externally.
Premiere Pro, Adobe’s video editing application, now includes a professional color grading system and an automated object removal feature for video clips. Lightroom, which manages photo libraries and basic editing, added automated correction for blinking eyes in photographs and a blemish removal tool that uses AI to detect and smooth skin imperfections.
Illustrator and InDesign also received AI enhancements as part of the broader rollout, though Adobe offered fewer details on the specific capabilities added to these vector and layout applications.

Model Context Protocol Opens Adobe Features to External Chatbots
Adobe is releasing a local Model Context Protocol (MCP) implementation that bridges Creative Cloud applications with external AI assistants. Users can now access selected Adobe features through ChatGPT, Claude, or Gemini without launching Adobe software. For instance, a user working in an AI chatbot could request specific edits to a Photoshop file or modifications to an Illustrator design and receive results without switching applications.
Deepa Subramaniam, an Adobe vice president, told TechCrunch that the company continues to evaluate which capabilities warrant exposure through MCP. We are still figuring out where and when and how to draw the line, Subramaniam explained. What we’re putting out there is a number of endpoints and capabilities across Illustrator, Photoshop, After Effects and Premiere. The company will use patterns in how users attempt to apply the protocol to inform which additional capabilities to surface next.
On the philosophy behind exposure, Subramaniam emphasized Adobe’s openness: We don’t have any need or desire to sort of limit what the MCP endpoint can expose. What we really want to do is to support that creative workflow, how you choose to do it. This positions the MCP strategy as workflow-centric rather than application-centric, recognizing that creative professionals increasingly work across multiple platforms and interfaces.
The MCP release builds on Adobe’s existing integrations with ChatGPT and Claude that grant users access to some product functions. The protocol approach promises deeper access and more granular control than the earlier API integrations offered.
Scrutiny From Competitors on MCP’s Neutrality
Last month, Figma co-founder Dylan Field raised concerns during an event in India about whether MCP truly benefits all companies equally. Field suggested that MCP’s architecture disproportionately advantages large frontier AI model developers over other vendors seeking to connect their tools. His skepticism reflected broader industry debate about whether open protocols become dominated by well-capitalized players.
Adobe’s decision to deploy its own local MCP server suggests the company is willing to participate in the standard despite reservations from some quarters, betting that the reach into environments where users already work outweighs concerns about the protocol’s design.
Topaz Labs Acquisition Brings Specialized Image Models
Adobe continues to layer third-party AI capabilities into its suite, already offering integrations with models from Google, Runway, ElevenLabs, Black Forest Labs, OpenAI, Kling, and ByteDance. The company’s recent acquisition of Topaz Labs brought image processing expertise in-house, and Topaz is now releasing two models within Adobe’s environment.
Wonder 4 handles image upscaling and denoising, addressing professional photographers’ need to improve photograph quality and reduce noise from high-ISO captures. ImageHDR converts standard dynamic range images into HDR format, expanding tonal range and detail in highlights and shadows. Both models ship within Adobe’s apps, reducing friction compared to managing external tools or services.
Source: TechCrunch
Frequently Asked Questions
What is the Adobe MCP and how does it work?
Adobe is releasing a local Model Context Protocol that connects Creative Cloud apps to external chatbots like ChatGPT, Claude, and Gemini, allowing users to access select Adobe features through those interfaces without launching Adobe software.
What new features did Firefly receive at Adobe MAX?
Firefly added Builder (a prototyping tool in private beta), graph generation, selective image editing, AI-powered video assembly, and became proactive to suggest actions based on active project context.
Which specific applications received updates at Adobe MAX?
Photoshop got relighting and generative styling; Premiere Pro gained color grading and object removal; Lightroom added eye blinking correction and blemish removal; Illustrator and InDesign also received AI enhancements.