Adobe MAX 2026 showed me the future of Photoshop, Premiere and Illustrator – and the wildest AI tools are still to come
Adobe is opening up Creative Cloud to third-party AI agents, but what does that mean for creative control and how can we trust the results?
At the company’s major annual conference in Miami, I watched an Adobe researcher take a painting created almost 400 years ago and move a virtual camera around parts of the scene that the artist never painted. What appeared in that demo was remarkable. It was also unsettling.
I saw some of the most experimental creative software presented in Sneaks, co-hosted by Terry Crews. I also spoke with Adobe’s top executives about the future of AI-powered workflows, what it means for creative professionals and the growing questions around authenticity and trust.
First off, let’s start with the most exciting highlights of the show, which were revealed during the famous MAX Sneaks session.
Sneaks is where Adobe researchers present projects that they’ve been working on. The range of experimental tools on display run from practical and fun, to mind-blowing witchcraft. Unfortunately, not all ideas presented here will make it into Adobe Creative Cloud, but occasionally, some do. The recently released Relight feature is one such project that was shared at MAX Sneaks 2025, which eventually shipped to Photoshop users before MAX 2026.
MAX Sneaks Magic
Imagine you captured a photo, but later want to change the angle and create a new image from a unique perspective. Project Sight Span transforms any single image into an explorable environment, letting you move a virtual camera around the scene to create entirely new perspectives.
Yiwei Hu, Senior Research Scientist at Adobe, started his presentation with an image of a panda laying across a branch. He changed the perspective from side view to front-on. Rotating 45° within the image. The preview was blotchy, but as soon as it was rendered, the detail of the original image was there.
He showed a couple more photographic examples that were equally impressive. But his final example was approaching Sci-Fi levels of crazy.
Get all the latest news, reviews, deals and buying guides on gorgeous tech, home and active products from the T3 experts
Yiwei Hu, pulled up a painting: The Milkmaid, painted by Dutch artist Johannes Vermeer c. 1658. He then proceeded to turn the virtual camera around within the scene, showing elements of the room the painting was set in and then tilted the camera above the milkmaid to create a new image, placing more emphasis on the contents of the table she was working at.
Once rendered, the model was able to convincingly extrapolate and depict details that were never in the original image, but in the same style. The possibilities of such technology are impressive and concerning, in equal measure.
Have you ever embarked on a long editing session that involved repetitive fixes that you wish could be automated? Well, Project Rinse Repeat could solve that and more.
Presented by Adobe Senior Computer Scientist Ridam Batra, this feature allows Photoshop to see what you’ve edited, copy it and then apply the intentions of your edits to other areas of the image. The thing I love most about this is that it uses AI to speed up your work, without removing creative input. All of its edits remain fully adjustable, as if you’d performed them yourself. As Ridam Batra put it in her presentation: “This does not take away the control from me, it only takes away the repetition.”
The best example she gave was a quick skin blemish edit on a group photo. She cleaned up one person’s skin, hit “Copy Edits”, then drew a box around the other three people in the image and hit “Paste Edits”, and Photoshop was able to understand the intent and cleaned up the remaining people’s faces, who had varying degrees of shiny patches and spots. The feature worked so well and will save photographers a huge amount of time if it’s ever deployed. I suspect it will be.
Finally, Project Deep Brush allows the brush and text tools in Photoshop to understand the scene geometry of surfaces, lighting and objects within the image.
The demo included painting on cloth material with curves, angled surfaces and around a person’s limbs as they pushed off a wall. The brush adapted to the shape and angle of whatever it hovered over in real time, as did text, and if you checked the warp function, it would immediately mould to the texture of the surface it was overlaid on. Another example of this tool’s power saw presenter Iliyan Georgiev, Adobe Senior Research Scientist, paint a neck tattoo on a portrait of himself and then add light into the scene using a brush, all the while respecting the scene’s existing geometry.
Of the ten demonstrations that took place during Sneaks, this was the most practical and the one I would put my money on coming to Photoshop first, especially as it’s already available for people to try at Labs.adobe.com/playground.
Pro Tip: Also check out Project Hands-On while you’re there.
Agentic editing in Creative Cloud
Unlike conventional generative AI, which responds to prompts by producing an image, text, video or sound, agentic AI can understand a goal and then plan and execute the steps required to deliver on your request using the software tools available to it.
MCP (Model Context Protocol) makes it possible for AI agents to operate other software. It’s not an Adobe invention or a new AI model, but an open standard that Adobe is adopting to make its Creative Cloud applications accessible to third-party-driven AI assistants.
Think of MCP as a bridge between AI tools like ChatGPT or Claude and software like Photoshop or Premiere. What will this allow us to do? Let’s say you recorded a podcast episode. You could ask the AI assistant in Premiere to identify strong clips and create a batch of highlight Reels for social media, and translate your text graphics and transcript for other markets. Instead of giving you tips for how to do those things, Adobe wants your AI assistant in its software to be able to perform those tasks for you.
Crucially, the resulting files aren’t compressed AI-generated final deliverables. They will open a new file, sequences or project, carry out the edits directly within your Creative Cloud software, leaving fully editable layers, timelines, vectors and elements. This is intended to give creatives full control over the final exports. If it works as Adobe described, we will be able to go into those AI-generated outputs and refine them in as much detail as if we had created them the “old-fashioned way” ourselves.
With private beta expected to open up in 2027, I don’t yet know what the full range of available commands will be, how permissions and approvals will work, or crucially, how this new workflow will be priced. And I think these finer details are where the rubber will really hit the road.
AI-friendly reception
When Adobe announced that MCP agentic workflows were coming to Creative Cloud, the room popped with glee. But speaking with other creatives, the response to AI in general has been tepid, if not downright cold. Many have justifiable concerns around the increasing prevalence of AI tools and AI-generated content flooding the workplace and social online spaces.
So what is it about Adobe’s captive audience that made this such a popular addition? I think that Adobe’s approach goes a long way toward disarming some people’s hesitance and disdain for AI. These tools are being presented as ways to remove the time and energy cost of repetitive work, not as a shortcut to creativity. Adobe’s executive team and its stream of talented keynote speakers were at pains to state that there’s no replacement for creativity and genuine talent.
One of my favourite moments of the entire event came from Siobhán Maroney, Creative Director and Co-Founder, They Are Creative. There were many brilliant takeaways, but one of the statements she projected across the massive screens behind her has stayed with me: “The only thing that’s real is your imagination.”
Given the raging debate around AI’s place in the creativity pipeline, I really appreciated the perspective. And it’s something I speak about frequently. AI alone does nothing; it needs human, creative thought and input to begin and human curation to craft its outputs into something coherent and impactful.
Where is this heading?
I can’t help but feel like Adobe’s ultimate strategic advantage could actually lie in its sophisticated, device-agnostic creative applications, rather than individual products like Photoshop.
As I reflect on Adobe MAX, I keep returning to Siobhán Maroney’s inspirational keynote. Siobhán focuses on three things: overdelivering for her clients, keeping up with the latest tools, and passionately embracing whatever is available to help bring her creative vision to life.
It’s actionable advice for anyone keen to make a success of being a creative professional. But as the tools that turn our imaginations into convincing images become more powerful, I think we need to become equally ambitious about preserving our ability to distinguish imagination from objective reality.
Jon Devo is a professional photographer and videographer of 20 years. He became a presenter and journalist after studying Journalism and Psychology at City University London in 2009. Jon founded the Devoted Creatives photo and video studio and is the technology & science host of Spotify’s Shot & Chaser podcast. Jon has a monthly column, called Scanning Ahead, and contributes for a number of leading publications and tech platforms, including T3.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.