Author: Merrel Briona

Adobe previewed 10 research projects at Adobe MAX 2026 Sneaks, spanning video, audio, photography, 3D, imaging and design.

The projects reflect the company’s ongoing research into creative tools and workflows. Some are also available for hands-on testing through the Adobe Labs Playground.

Video tools

Four previews focused on video creation and editing.

Project Scene Sweep is aimed at editors working across multiple shots in a sequence. It identifies characters, locations and scenes so an edit can be applied consistently across cuts and camera angles, rather than repeated manually.

Another project, Screen Test, focuses on character development. It lets creators build a fictional character’s appearance, voice and personality, then use natural language dialogue to explore how that character might respond in different scenes.

Live Set explores real-time changes within a moving scene. Users can alter a subject, change an environment or redirect the camera while a video continues to play.

On Cue addresses the timing of events in generated video. Creators can direct when an action takes place, such as a flash of light, using an editable timing signal drawn by hand or extracted from existing video or music.

Audio work

Adobe also highlighted a project focused on sound editing.

Sound Bender is designed to pull a described sound from a finished mix and let users explore variations on an interactive canvas. Creators can then shape a sound in real time, align its movement with video or generate a new sound from a text description.

Image editing

Five previews focused on images and design workflows.

Deep Brush uses AI to interpret the shape, colour and lighting of an image. This allows brushstrokes to follow surfaces, enables colour changes that preserve shadows and lets users paint light directly into a scene.

Sight Span turns a single image into a navigable environment, letting creators move within the scene, find new angles and compose shots beyond the original frame.

Hands On uses object segmentation and vision-language models to place controls directly on the image canvas. This lets users manipulate elements of a scene by hand rather than relying solely on menus or separate controls.

Rinse Repeat is intended to copy edits from one part of an image to another. The process recreates those edits with native Photoshop tools so they remain editable for later refinement.

Link Up explores the use of gesture in design work. A drag action can express a creator’s intent, such as placing a car in an illustrated parking space, after which the system adjusts perspective, scale and style to match the surrounding scene and leaves the result as editable vector artwork.

Labs access

Alongside the stage previews, Adobe expanded Adobe Labs as a channel for ongoing research feedback.

Adobe Labs now combines research updates, professional workflows and early-stage experiments from across the company’s teams. Within that broader offering, the new Playground gives users access to selected previews, including Deep Brush and Hands On, as well as a Photoshop feature preview called Custom Filters.

The move points to a wider effort to put some experimental work in front of users earlier, rather than limiting exposure to annual demonstrations. It also gives Adobe a way to gather feedback from creative professionals and hobbyists as it assesses which ideas might develop into product features.

For Adobe, the Sneaks format has long served as a public window into research that may later influence commercial software. This year’s projects continued to emphasise automation paired with direct user control, especially in areas such as scene continuity, timing, image manipulation and sound extraction.

The 10 projects also indicate where Adobe sees persistent friction in creative work: repeating edits across shots, isolating audio from mixed material, transferring image adjustments and making design changes more directly on the canvas.

Within Adobe Labs, the new Playground lets users try selected early innovations themselves, including Project Deep Brush and Project Hands On, and preview Project Custom Filters for Photoshop.

The global rollout, announced Wednesday (Nov. 11), brings Gemini Live directly to CarPlay, allowing drivers to hold natural, back-and-forth conversations with the assistant through their vehicle’s display. Users can start a new conversation or resume a recent one by tapping the Gemini icon on the CarPlay screen.

Google said the integration is designed to let drivers accomplish more during their commutes while minimizing distractions. Users can interrupt Gemini to provide additional information, change subjects or pause conversations when they need to concentrate on driving.

The integration also extends Gemini’s reach into Google’s productivity ecosystem. Through connected Google Workspace applications, including Gmail and Google Calendar, drivers can ask Gemini to summarize unread emails, review long message threads, check upcoming appointments and prepare for meetings.

The rollout is available globally to Gemini users on iOS in all languages supported by Gemini Live. To access the feature, users must update the Gemini application on their iPhones, enable microphone permissions and have used Gemini Live on their phones at least once.

The move expands Google’s competition in the growing market for conversational AI assistants across mobile and connected devices. Gemini already supports Android Auto and vehicles equipped with Google built-in, while its earlier CarPlay integration was limited to Google Maps.

Gemini’s arrival also follows the expansion of other AI assistants onto CarPlay, including OpenAI’s ChatGPT, Anthropic’s Claude, Microsoft’s Copilot, Perplexity and xAI’s Grok. The additions give drivers more ways to access conversational AI through their vehicles rather than relying exclusively on traditional voice assistants.

The expansion builds on Google’s broader push to make Gemini a conversational interface across consumer and enterprise applications. Google said in September that its Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking models were designed to power more capable voice agents, with improvements in real-time reasoning and complex task completion.

The Gemini app surpassed 1 billion monthly active users in August, becoming Google’s fastest-growing product to reach that milestone, according to the company.