Adobe's experimental camera app now critiques your photos with AI-powered feedback

Adobe has expanded its experimental iOS photography app, Project Indigo, with AI-driven capabilities designed to analyze and suggest improvements to user photos. The new tools leverage language models to provide constructive feedback and automated editing options, moving beyond simple prompt-based interfaces.

Photo critique and capture guidance#

The core additions include two recommendation features. A professional photo critique examines framing, lighting, color balance, and emotional resonance, offering detailed assessment rather than generic tips. A separate capture-and-edit suggestion mode advises users on how to reshoot a scene or adjust composition, exposure, and subject matter—for example, recommending removal of specific elements cluttering the frame.

Rather than requiring users to craft precise text prompts, these tools function as toggle-based controls generating predictable outputs. Marc Levoy, leading the project after his prior work on Google Pixel's camera systems, explained the rationale: prompt engineering often proves cumbersome, so the team prioritized deterministic features that function more like buttons than open-ended text interfaces.

Advanced editing capabilities#

Project Indigo now includes intelligent object removal using toggles for common unwanted elements—people, trash, power lines, vehicles, and fences—plus custom object description. Early tests show the feature removes subjects cleanly without artifacts. The app also applies AI-generated depth of field to simulate shallow-focus backgrounds on flat photos.

A style transfer mode lets users rerender images as watercolor, pen-and-ink, monochromatic, or backlit scenes. While reminiscent of early filter apps, the results appear more polished with advanced models, though the feature is not functionally novel.

Rollout and model choices#

Currently powered by Google's Gemini Nano Banana model, Adobe remains open to integrating alternatives, including its own Firefly engine. These capabilities sit within an "AI playground" section still undergoing limited testing; the company has not committed to broader release.

Levoy acknowledged the tension between his preference for deterministic controls and Project Indigo's own freeform edit-description feature—a flexibility that risks producing lower-quality outputs. Still, the suite aims to educate photographers on compositional nuance rather than simply automating mediocre edits.