Adobe Research

With a team of world-class research scientists, engineers, artists, and designers, Adobe Research combines cutting-edge academic discovery with industry impact.

[Adobe Research @AdobeResearch]

A/B testing typically takes weeks and significant budget. Project Face Off from Adobe Research collapses that to seconds. AI personas with distinct demographics, opinions, and personalities predict which designs will perform best — and explain why. #AdobeSummit Sneaks voters agreed: it won best Sneak of the year. https://t.co/LevZh0cEYg

[Adobe Research @AdobeResearch]

Adobe Research helped build Photoshop's new AI Assistant, nearly a decade in the making. The core insight from user research: some people want the agent to handle it all, others want to control the process and learn. So the team built both. Now in beta on web and mobile. https://t.co/tLgvUwJewH #AdobeResearch #Photoshop

[Adobe Research @AdobeResearch]

Adobe Firefly’s alt text feature is helping blind and low-vision creators work with AI-generated images in new ways — making creative workflows more accessible.

Learn more: https://t.co/9dS0ce6JHz

#Accessibility #AdobeFirefly

[Adobe Research @AdobeResearch]

How do you reward an AI for creative editing when there’s no single right answer? RetouchIQ introduces a generalist reward model that generates tailored evaluation criteria per edit. New paper from Adobe Research & UC Santa Barbara. https://t.co/v5yHF4hJp9 #CVPR2026 #ComputerVision #MLLM

[Adobe Research @AdobeResearch]

Get a peek at some of the 75+ cutting-edge papers Adobe is sharing at #CVPR2026. #AI #ComputerVision https://t.co/gYOlTXXky9

[Adobe Research @AdobeResearch]

How do you make AI-generated 3D shapes editable? Break them into smart building blocks. Our #CVPR2026 Oral SuperFit converts 3D shapes into compact, editable primitive assemblies with SuperFrustum, an 8-parameter primitive that spans cubes, cones & more. https://t.co/7bqiVSISM9 #3D

[Adobe Research @AdobeResearch]

What if every image editing instruction had a slider? SliderEdit, new from Adobe Research at #CVPR2026 (Oral) turns image editing instructions into adjustable sliders, enabling smooth control from subtle changes to dramatic transformations. Unlike prior approaches, SliderEdit uses a single lightweight framework that generalizes across diverse edits and prompts.

Read more: https://t.co/b2ME14113p

[Adobe Research @AdobeResearch]

Adobe Research is proud to co-organize a CVPR 2026 tutorial on June 4 on watermarking and trustworthy generative AI, alongside collaborators from Meta, UMD, and University of Siena. From classical foundations to multimodal and generative watermarking, benchmarking, and deployment, we'll explore the full stack of media authenticity. Check out the tutorial website for more details: https://t.co/aWCCNsAj6Q

#CVPR2026 #AI #ContentAuthenticity

[Adobe Research @AdobeResearch]

🔍 Adobe Research is co-organizing the Authenticity & Provenance in the Age of Generative AI (APAI) workshop at #CVPR2026!

�� June 3 | 1–6 PM | Denver, CO
🎤 Speakers from Google DeepMind, FAIR, DSRI & WITNESS
🏆 Video & Audio Splice Detection Challenge

Check out 👉 https://t.co/Yg7zKxUgi5
#AdobeResearch #AI #MediaForensics #ContentAuthenticity

[Adobe Research @AdobeResearch]

New experimental work from Adobe Research for ACL 2026: TeamFusion creates AI proxy agents grounded in each designer’s preferences, runs a structured debate, and synthesizes consensus. Designers largely agreed with their proxies and reached stronger consensuses.

📄 https://t.co/fkbC7iqHYY #NLProc #ACL2026NLP

[Adobe Research @AdobeResearch]

What if AI could review your design like an expert panel? 🤖 Agentic Design Review System from Adobe Research (AAAI 2026) is an experimental framework where agents evaluate typography, color, alignment, and coordinate actionable feedback. 📄 https://t.co/4IcbdzZyZC #AdobeResearch #AAAI2026 #AIAgents

[Adobe Research @AdobeResearch]

A/B tests can be slow and expensive, but an experimental tool from Adobe Research may change that. Project Face Off gives users a team of AI personas that predict which design ideas will hit marketing goals and share actionable suggestions to make campaigns even more compelling. Check out a Sneak of Project Face Off from #AdobeSummit 2026 - voted Best Sneak! https://t.co/saTsL6Hbsn

[Adobe Research @AdobeResearch]

What if designing with #AI felt as fluid as a real brainstorm? Project Test Kitchen, an in-development tool from Adobe Research, lets users combine text prompts, reference images, and sketches to generate options across a range of styles. Pick your favorite elements, recombine, and keep iterating. See a Sneak of Project Test Kitchen from #AdobeSummit 2026! https://t.co/8XV0X4JZBb

[Adobe Research @AdobeResearch]

Infographics that never go stale? That's Project Concurrent. This experimental tool from Adobe Research links infographics directly to live data, so they update automatically. No manual refreshes needed. Here's a sneak preview from #AdobeSummit 2026. https://t.co/DBL61pnSxZ

[Adobe Research @AdobeResearch]

At this year's #AdobeSummit, Adobe Researchers were behind some of the biggest announcements. They shared new #AI agents for orchestrating customer experiences at scale, plus Sneaks of tools that let you design alongside AI. Think: AI personas that tell you which campaigns will actually convert, or infographics that update themselves. Check it out! https://t.co/xnd0Bi6YsT

[Adobe Research @AdobeResearch]

Remove the object. Remove the shadow. Remove the reflection. 🎬 Object-WIPER is an experimental training-free method for erasing objects and their visual effects from video. No retraining, no frame-by-frame cleanup. Accepted to #CVPR2026 🎉 Read more: https://t.co/vYgBA28aLi
#ComputerVision #VideoEditing

[Adobe Research @AdobeResearch]

Put any subject, anywhere, in one shot.

SIGMA-GEN from Adobe Research (ICLR 2026) is the first method to preserve multiple identities AND respect spatial constraints — 2D boxes, 3D boxes, masks, depth — in a single generation pass. New dataset: SIGMA-SET27K (100K+ subjects). https://t.co/B6qyWpMFpW

#AdobeResearch #ICLR2026 #ComputerVision

[Adobe Research @AdobeResearch]

Imagine controlling what happens moment by moment in your AI-generated video, at the same time as the video is being created. Adobe researchers are making this possible with their experimental technology MotionStream. https://t.co/ZZN3chdQX9

[Adobe Research @AdobeResearch]

Your footage has gaps. Vidmento fills them. Experimental work from Adobe Research at #CHI2026 : an AI video authoring tool that blends your captured footage with contextually generated clips — matching style, tone, and story. Scaffold, don't automate. Creators stay in control. 🎬
Paper: https://t.co/BHd6qnKLbp
#CreativeAI #VideoAI

[Adobe Research @AdobeResearch]

New at #CHI2026: VidTune helps video creators find the right soundtrack — not by listening to every track, but by seeing them. The system generates music from a prompt and produces contextual visual thumbnails that encode mood, energy, genre — grounded in your video's imagery. From @AdobeResearch and @UCBerkeley https://t.co/ibMyi8EsKu

#HCI #GenerativeAI #CreativeAI