Adobe AI now judges your photos

Adobe is testing new AI-powered tools in its experimental iPhone camera app, Project Indigo, that review photos and recommend edits instantly. The features, currently in limited testing, use large language models to assess framing, lighting, and composition before offering specific adjustments.

The app introduces two feedback features. One evaluates images against professional standards, checking elements like lighting, color balance, and emotional impact. The other provides capture and edit suggestions, such as adjusting exposure or removing distractions before taking another shot.

During a test, the app identified a hexagonal white object and recommended removing it. It also proposed Lightroom-style adjustments, like increasing contrast or correcting white balance, to enhance existing photos.

Marc Levoy, who leads Project Indigo and previously worked on Google’s Pixel camera, noted that most generative AI tools depend on text prompts, which can slow down quick edits. Project Indigo instead uses preset buttons to deliver more consistent results.

This method contrasts with Google’s Pixel camera coach, which gives general framing advice. Project Indigo’s suggestions are more precise, aiming to help users learn photography techniques even if they don’t always agree with the AI’s advice.

Adobe has also improved object removal, a feature available in Photoshop and Google Photos. Traditionally, users had to manually trace objects, often leaving visible flaws. Project Indigo simplifies this with toggles for common distractions like people, trash cans, wires, fences, and vehicles.

Users can also describe custom objects to remove. In one test, the app erased a person and the item they held without obvious errors. The outcome was smoother than many existing tools, though Adobe calls the feature experimental.

The app now includes AI-generated depth of field, creating a blurred background to highlight subjects. It also offers style transfer options, applying filters like watercolor or ink line effects. While similar tools exist in apps like Prisma, Adobe’s version produces more refined results.

Despite Levoy’s doubts about prompt-based editing, Project Indigo includes a text-to-edit function. Users can request changes, such as describing a custom edit, but Adobe cautions this may lead to unnatural results. The app currently uses Google’s Gemini Nano model but may later switch to Adobe’s own Firefly AI.

For now, the AI tools appear under an “AI playground” tab and are limited to select testers. Adobe hasn’t confirmed broader availability, but the project shows how AI could make advanced photography easier without requiring expertise in complex software.

The balance between helpful guidance and excessive editing remains delicate. The critique feature might encourage formulaic compositions, while style transfer could turn photos into generic digital effects. The value of these tools depends on how users apply them.

These updates reflect Adobe’s efforts to integrate AI into creative tools, though the company has avoided overpromising. Project Indigo debuted last year with manual controls and multi-frame super-resolution, serving as a testing ground for both amateurs and professionals.

Levoy’s group continues refining the AI features, and some may remain in testing indefinitely. For now, they provide a rare case of AI being used to educate, not just automate, photography skills.