Adobe camera app’s new feature will critique your photos using AI

Adobe is pushing its experimental iOS camera app, Project Indigo, further into AI territory. The app, which launched last year with pro controls, multi-frame super-resolution and various capture modes, is picking up a set of features that use large language models to actually tell you what's wrong with your shot. Marc Levoy, who leads the project and previously shaped the camera smarts on Google's Pixel phones, is the driving force behind the approach.

Levoy's argument is straightforward: prompt-based AI editing tools are annoying because finding the exact wording to get the result you want is often a guessing game. So instead of another prompt box, most of Indigo's new tools are buttons that spit out more predictable, deterministic results. That's a different philosophy than Google's own camera coach feature for Pixel, launched last year, which sticks to fairly generic framing tips. Indigo's suggestions go further, offering specific, descriptive feedback that could actually teach someone something about photography, even if they end up ignoring the AI's opinion.

There are two pieces to this critique system. One gives you a "professional" style breakdown covering framing, lighting, color and emotional impact. The other offers capture and edit suggestions, pointing out things you might reshoot differently, like tweaking exposure or removing an object from the frame entirely. One tester was told to get rid of a hexagonal white object cluttering a shot. A separate section takes an existing photo and recommends how to improve it using Lightroom's editing controls.

Object removal itself isn't new. Apple Photos, Google Photos and Photoshop have all offered it for years, usually requiring you to circle or select the offending item by hand, with mixed results. Indigo instead gives toggles for common annoyances: people lurking in the background, trash and trash cans, wires, poles, fences, vehicles, general clutter, or a custom object you type out yourself. In testing, it managed to erase a friend and the object he was holding from a photo without leaving obvious artifacts behind, which is more than can be said for a lot of automated cleanup tools.

The update also adds AI-generated depth of field to fake a blurred background, plus a style transfer feature that can render a photo as watercolor, pen and ink, ink line with color wash, monochromatic or backlit. The results look polished, though the whole concept feels like a callback to the early Prisma app rather than anything genuinely new. And despite Levoy's stated skepticism toward prompt-driven editing, Indigo quietly includes its own version, letting users type out edits that presets can't handle. That's arguably the feature most likely to produce the kind of AI slop Levoy seems eager to avoid elsewhere.

Under the hood, these tools currently run on Google's Gemini-based Nano Banana model, though Adobe says it could swap in other models, including its own Firefly, down the line. Everything lives inside an AI playground tab and remains in testing, limited to a small group of users. It might never graduate to a wider release. But if it nudges people toward understanding why a photo works or doesn't, rather than just handing them another one-tap filter, that alone sets it apart from most of what passes for AI photography features right now.