Adobe's Project Indigo Adds LLM-Powered Photo Critique and Advanced Object Removal

According to TechCrunch, Adobe is extending Project Indigo, its experimental iOS camera application launched in 2025, with large language model-powered photo analysis and advanced AI editing capabilities. The update introduces LLM-driven critique that assesses composition, lighting, palette, and emotional resonance, alongside deterministic AI features including object removal, depth-of-field synthesis, and style transfer effects.

LLM-Based Photo Critique as a Teaching Tool

Project Indigo’s photo critique feature delivers assessments framed as “professional” opinions on photographic fundamentals. Unlike prompt-based generative systems that require users to craft specific requests, TechCrunch reports that this critique operates as a preset function—one tap triggers a full analysis without parameter tuning. Marc Levoy, who previously led computational photography for Google Pixel’s camera system, framed this design choice as a response to friction in typical generative AI workflows. The feature aims to educate photographers on technique rather than simply optimize aesthetics, even when users disagree with the assessment.

A secondary suggestion mode, capture and edit recommendations, guides users on reshooting angles or applying Lightroom adjustments to strengthen existing photos. The app identifies specific framings, exposures, or foreground elements that could be modified—and provides concrete next steps within the Adobe editing ecosystem.

Advanced Object Removal and Depth Simulation

Project Indigo’s object removal tool represents a departure from selection-based prior art in Apple Photos, Google Photos, and Adobe Photoshop. Rather than requiring manual lasso or brush-stroke selections, the app presents toggles for common removal targets: background subjects, refuse, cables, barriers, and vehicles. Users can also input custom descriptions for less common objects. TechCrunch’s testing found the results “impressive,” with clean removal of background figures without visible artifacts.

The app also synthesizes artificial depth-of-field, simulating shallow focus effects, and applies style transfer using models trained on watercolor, pen-and-ink, monochromatic, and backlighting aesthetic presets. TechCrunch notes these style outputs resemble early-stage Prisma output but with refinement from more advanced generative models.

Why This Matters

Adobe is positioning Project Indigo to compete in on-device camera AI, a space increasingly dominated by Apple and Google hardware. By embedding LLM-driven critique and deterministic editing as buttons rather than prompt fields, Adobe reduces user friction and targets both casual photographers seeking real-time guidance and enthusiasts wanting Lightroom-grade control. The object removal quality benchmarked by TechCrunch suggests Adobe’s computational photography stack has closed the gap with Pixel’s physics-based approaches—a meaningful signal for mobile editing market share dynamics.