Adobe's Project Indigo camera app adds AI photo critiques

Adobe is expanding Project Indigo, its experimental iOS camera app launched last year, with a set of AI-powered features built on large language models (LLMs). The app already offered pro controls, multi-frame super-resolution, and multiple capture modes. The new additions include a photo-critique tool, edit suggestions, advanced object removal, AI-generated depth of field, and style transfer.

Marc Levoy, who heads the project and previously developed camera technology for Google’s Pixel phones, said most generative AI editing tools rely on prompt-based interfaces, where finding the right prompt to get a desired result can be tricky. That is why most of the new experimental features are built as buttons that produce more deterministic outputs rather than open-ended prompts.

Two features fall into this category. The first, photo critique, offers a “professional” opinion on a photo’s framing, lighting, colors, and emotional impact. The second, capture and edit suggestions, gives tips on how to reshoot a photo — adjusting framing, exposure, or removing objects from the viewfinder. A separate section suggests how an existing photo could be improved using Adobe Lightroom controls.

By comparison, Google’s camera coach feature for Pixel phones, also launched last year, offers more generic framing suggestions; Project Indigo’s feedback is more descriptive and could help users learn photography techniques, even if they disagree with the AI’s assessment.

Object removal has existed in apps like Apple Photos, Google Photos, and Adobe Photoshop for years, but typically requires users to manually select the object. Project Indigo’s version adds toggles for common clutter — people in the background, trash and trash cans, wires and poles, fences, and vehicles — plus the option to describe a custom object to remove. In TechCrunch’s testing, the tool removed a person and the object they were holding from a photo’s background without introducing visible artifacts. The app can also simulate a blurred background to create depth of field, and apply style transfers such as watercolor, pen and ink, and monochromatic looks; TechCrunch’s testing found those style outputs refined but not a novel capability compared with the critique and object-removal tools.

Despite Levoy’s stated preference for deterministic tools over prompt-based editing, Project Indigo also includes a feature that lets users describe an edit in their own words.

Adobe is currently using Google’s Gemini-based Nano Banana model to power these features but says it is open to swapping in other models, including its own Adobe Firefly. The tools are bundled under an “AI playground” tab and remain in a testing phase, available only to select users. Adobe says the features might never reach a wider audience.

Why it matters

Pairing generative AI with fixed, button-based outputs instead of open prompts is Adobe’s bet on making AI photo editing more predictable as rivals like Google add similar camera-coaching features.