Opinion: Adobe's Project Indigo Experiments With AI Photo Critiques and Precision Cleanup
AI-generated image · Bay Street Wire
The latest updates to Adobe's experimental iOS app move beyond simple prompts to offer deterministic AI tools and professional-style composition feedback.
I've always believed that the best shot comes from the raw glass and a trained eye, so the idea of an algorithm critiquing my composition feels like a bridge too far. However, from a hardware and software integration standpoint, Adobe's latest moves with its experimental iOS camera app, Project Indigo, are worth a closer look.
According to TechCrunch, Adobe is introducing LLM-powered features to Project Indigo that allow the app to critique photos and offer editing suggestions. Marc Levoy, who leads the project and previously developed the camera features for Google's Pixel, noted that prompt-based editing can be a difficult process for users trying to achieve a specific result. To solve this, Levoy is implementing buttons designed for more deterministic outputs.
Two specific AI-driven feedback tools are being tested, per TechCrunch:
* **Photo Critique:** This provides a "professional" assessment focusing on colors, lighting, framing, and emotional impact. * **Capture and Edit Suggestions:** This tool offers tips for reshooting a scene—including changes to exposure, framing, and the removal of specific objects in the viewfinder—as well as advice on how to improve an existing image using Adobe Lightroom controls.
While Google launched a "camera coach" for Pixel phones last year that offered generic framing tips, TechCrunch reports that Project Indigo's suggestions are more descriptive.
On the utility side, I'm more interested in the cleanup. While object removal is standard in apps like Google Photos and Apple Photos, it usually requires manual circling. TechCrunch reports that Project Indigo introduces toggles to automatically remove specific clutter, such as vehicles, fences, poles, wires, trash cans, and people in the background. Users can also describe custom objects for removal. In testing conducted by TechCrunch, the app successfully removed a person and an object they were holding without creating strange artifacts.
Other additions to the "AI playground" tab include depth of field generation to simulate blurred backgrounds and style transfer options—such as monochromatic, watercolor, and pen and ink—though TechCrunch notes that style transfer is not a particularly new or useful feature.
Technically, Adobe is currently utilizing Google's Gemini-based Nano Banana for these features, but TechCrunch reports the company is open to using other models, including its own Adobe Firefly. These tools are currently in a testing phase and available only to select users, meaning they may never see a wide release.