If you've ever tried to use AI image generation in a real creative workflow, you'll know this particular misery. You spend time tweaking your prompts, refining a character's look, and finally get what you were looking for. You then ask the tool to show her in a different pose… but get back someone who looks like a distant relative at best.
This has been one of the most stubborn practical limitations of generative AI, and it's the thing that's stopped it from being useful for anyone wanting to make client-ready mockups or storyboards for campaigns, animation, comics or any other visual discipline requiring continuity.
Now, though, Adobe reckons it's cracked it. At a press briefing this week, the company showed off a suite of new features across Firefly and Creative Cloud (the company's second set of announcements this week – read about the first time-saving set here), but the one that really got my attention was Elements. It basically lets you define characters, locations and objects and reference them across generations of images and video alike.
Making Carla consistent
A demo by Paul Trani, Adobe's senior creative cloud evangelist, centred on a character called Carla, a 3D-animated girl with brown hair and a pink dress, created in Firefly using a single text prompt. With Elements, Carla can now be saved as a reusable reference and called up in any subsequent generation, maintaining her appearance across different poses, scenes and angles.
Paul showed off the feature live to us gathered journalists, conjuring Carla into a bedroom scene with a "fun, cute little alien cat" using barely more than her name and a brief description. Perhaps aware of the widespread antipathy to using gen AI for finished work, he noted: "I'm doing this for my girlfriend's niece."
Alongside Elements, Adobe is introducing Projects, a way to keep all your generations, assets and creative context together in one place. The combination of the two means you can build a story or campaign over time without losing your place or your characters.
Deepa Subramaniam, Adobe's vice president of product marketing for creative professionals, described both as foundational. "This is how you get consistency in your generations, whether you're generating images, videos, or both," she said. Both Elements and Projects are currently in private beta.
The agent arrives in Photoshop
The other major announcement is the public beta launch of Adobe's AI Assistant across Premiere, Photoshop, Illustrator, InDesign and Frame.io, with After Effects following in private beta. Rather than a generic chatbot bolted on, each integration is designed to respond to what creatives actually do in each tool.
Paul demonstrated the Premiere integration by asking the assistant to organise a project that was in absolute chaos, something many video editors and filmmakers will be familiar with. The assistant analysed the clips, renamed files descriptively and sorted everything into folders, including syncing multi-camera shots into a single composition. In Photoshop, meanwhile, the assistant fact-checked a design file and caught a date error, flagging that June 17th was a Wednesday, not a Thursday.
In Illustrator, Paul showed AI duplicating a circle 100 times with randomised positions, varied colours and size changes based on stacking order. "Could you imagine doing that yourself?" he asked. The Illustrator integration can also generate 50 versioned files from a spreadsheet and run pre-flight checks to catch colour mode errors or missing fonts before anything goes to print.
Deepa frames the underlying philosophy clearly in a blog post published alongside the announcement. "Creatives remain in control, choosing what to hand off, what to refine and how to apply your taste, expertise and judgment to shape every editable outcome," she writes. "These tools are built for how you've told us you actually work." It's a reassurance that will resonate with many: according to Adobe's own Creators' Toolkit Report, which surveyed more than 16,000 creators globally, 75% describe AI as integrated or essential to how they work, but 85% also say the final creative decision should always remain theirs.
Adobe has also added new agentic skills to the Firefly AI Assistant, aimed at social creators and smaller businesses. These include brand kit creation, short product video reels, Quick Cut for assembling footage automatically, and Storyboards for sequencing ideas before generating video.
During the brand kit demo, Paul prompted the assistant to build a brand for a jewellery business called Ola Jewelry, which then walked through questions about aesthetic direction and audience before generating logo concepts, colour palettes and typography, all editable and exportable as vector files. "Where's your vector of the logo?" he joked, in a nod to every small business owner who's only ever had a JPEG.
Finally, Adobe confirmed its creative agent is now available inside ChatGPT, Claude, Copilot and Slack, with Gemini support coming soon. The aim is to bring Adobe's tools to wherever creatives already work, rather than requiring them to leave their current environment.
For hardware that can handle it, see our best laptops for AI and best laptops for Photoshop guides.
