Adobe's Project Indigo now uses AI to critique your photos

Adobe is giving its experimental iOS camera app, Project Indigo, a brain — specifically, a large language model that judges your photography. Led by Marc Levoy, the engineer who built Pixel’s computational photography prowess, the app is moving beyond simple capture tricks into offering genuine, descriptive feedback on framing, lighting, and emotional impact. It’s a deliberate pivot away from the prompt-based editing Levi sees as a tricky, often frustrating user experience. Instead of typing magic words, you get buttons that produce more predictable results.

The new “photo critique” feature delivers a written assessment that feels like a mini photo class, pointing out what works and what doesn’t. Complementing this is a capture and edit suggestion tool that gets surprisingly specific — during a test, the app recommended removing a hexagonal white object from the frame. A second section then details how to fix issues using Lightroom controls. It’s far more granular than the generic framing tips Google’s Pixel Camera Coach offers, and could actually teach users a few things, even if they ultimately disagree with the AI’s taste.

Beyond coaching, Indigo is getting a handful of practical AI-driven editing tools. Object removal has been rethought: instead of painstakingly circling an unwanted person or wire, you now get simple toggles for common clutter — background people, trash cans, fences, vehicles. You can also describe a custom object to wipe away, and the results, in early testing, are clean, removing a person and the object they held without introducing weird visual artifacts. The app can also generate artificial depth of field for a blurred background effect and experiment with style transfer, turning photos into watercolor or pen-and-ink looks, which feels a bit like a polished throwback to the Prisma era.

Under the hood, Adobe is currently using Google’s Gemini-based Nano Banana model but is keeping the door open to its own Firefly model. All these tools are bundled in an “AI playground” tab and remain strictly in testing, with no guarantee of a public launch. The real intrigue isn’t just the feature set, but the philosophy: teaching people the nuances of photography rather than simply generating more AI slop. Levoy’s team is betting that constructive criticism, not just a limitless prompt box, is the more useful AI superpower for a camera app.