Adobe showcased a suite of AI-driven automation tools at the Adobe MAX 2026 conference in Miami, aiming to streamline complex image and video workflows.
By integrating large language models directly into the Creative Cloud ecosystem, Adobe plans to replace manual tool selection with conversational prompts for tasks like object removal and lighting adjustment.
Conversational Control Through Model Context Protocol
The core of Adobe’s strategy involves bridging the gap between professional-grade software and generative AI. At the conference, the company demonstrated how users can utilize interfaces like ChatGPT, Claude, and Gemini to manipulate complex assets within Photoshop, Premiere Pro, and After Effects. This is achieved via the Model Context Protocol (MCP), a technical interface that allows AI assistants to interact with local software files.
Unlike cloud-only generators, these integrations aim to maintain the structural integrity of projects. Changes are applied directly to the project files, preserving layers and video clips for further manual adjustment. According to Netzwelt, while initial connections to external AI services are already possible, the deeper local integration for these professional applications is slated for a closed beta release in early 2027.
Automated Restoration in Photoshop and Lightroom
Adobe is targeting the common frustrations of post-production with new, specialized features. The “Relight” tool in Photoshop allows users to place up to three virtual light sources into a scene, enabling targeted adjustments to shadows and exposure without affecting the entire image. This functions as a more granular alternative to global brightness sliders.
Lightroom is receiving an “Open Eyes” feature, designed to correct blink-related issues in group photos by pulling data from other frames of the same subject. This tool is currently in the testing phase. "Generative Restyle" also allows users to alter the aesthetic of an image using text-based descriptors, moving beyond simple object removal into stylistic transformation.
Firefly Evolves into a Creative Agent Managing Multi-Step Tasks
Firefly is evolving from a simple image generator into a broader “Creative Agent” capable of managing multi-step tasks. In a demonstration, Adobe showed how a user could request an Instagram post from a set of raw vacation photos. The AI handles the selection of images, performs edits, and generates a cohesive layout.
To prevent the AI from acting autonomously without oversight, Adobe introduced the “Plan Mode.” This feature provides a transparent look at the steps the AI intends to take, allowing the user to review or modify the sequence before execution. The platform is also opening up to third-party models from providers like Google and OpenAI, moving toward a more modular, model-agnostic architecture for the Firefly suite.
Premiere Pro Integrates Generative Object Removal and Audio Tools
Premiere Pro is integrating “Generative Remove” to handle the time-intensive process of cleaning up video footage. The tool identifies unwanted objects, such as background clutter, and fills the space using temporal data, improving upon static image-based object removal. This feature is currently in beta.
Audio processing is also receiving a major overhaul. New tools aim to isolate specific voices from background noise, such as traffic in an interview, and allow for individual manipulation of sound sources. The introduction of a “Creative Agent” for video suggests that Adobe intends to automate entire workflows in the near future.
Experimental Features in Adobe Labs
The “Project Deep Brush” technology represents a departure from traditional layer-based editing. Users can directly paint over areas of an image, with the AI calculating lighting and perspective to ensure the modifications remain consistent with the original scene. This experimental tool is currently accessible via the Adobe Labs Playground.
Other research projects, or “Sneaks,” include “Project Sight Span” for spatial environment creation, “Project Live Set” for dynamic video scene alteration, and “Project Sound Bender” for granular audio manipulation. These projects remain in the research phase. As noted by Netzwelt, there is no confirmed timeline for when or if these specific experimental features will be integrated into the final Creative Cloud software.
The Shift Toward Automated Complexity
Users should be aware that intensive use of generative functions is governed by “Generative Credits,” with the standard tier including 4,000 credits per month. Additional costs may apply for high-volume workflows, according to information provided at the event.
