Key facts
- Adobe's Project Indigo app is introducing AI-driven photo critique and editing suggestions.
- The app uses large language models to analyze framing, lighting, colors, and emotional impact.
- New AI tools include advanced object removal, depth of field generation, and style transfer.
- Features aim for deterministic outputs, unlike typical prompt-based generative AI.
- The AI features are powered by Google's Gemini Nano and are currently in testing.
Adobe is enhancing its experimental iOS camera application, Project Indigo, with new artificial intelligence capabilities designed to critique photos and offer editing guidance. Launched last year, the app previously provided professional controls and advanced capture modes. The latest additions leverage large language models (LLMs) to analyze images and suggest improvements.
Marc Levoy, who leads the project and previously contributed to Google Pixel's camera technology, noted that many generative AI tools rely on prompt-based editing, which can be difficult to perfect. To address this, Project Indigo's new AI features are designed as button-activated tools for more predictable results.
Key new features include a "professional" photo critique that assesses framing, lighting, colors, and emotional impact. Additionally, capture and edit suggestions offer real-time advice on adjusting framing, exposure, and removing unwanted objects within the viewfinder. The app also allows users to enhance existing photos using Lightroom-like controls.
While object removal has been available in other photo apps, Project Indigo's AI-powered version offers toggles for common clutter like people, trash, wires, and fences, and can also handle custom object removal with impressive results, avoiding artifacts. The app can also generate depth of field to simulate background blur.
Adobe is also experimenting with style transfer, enabling users to apply artistic tones such as watercolor or pen and ink to their images. Despite Levoy's critique of prompt-based editing, Project Indigo includes a feature for descriptive editing to perform custom edit flows.
These new AI functionalities are powered by Google's Gemini Nano model, though Adobe is open to integrating other models, including its own Firefly. Currently available only to select users in a testing phase under the "AI playground" tab, these features may not reach a wider audience but aim to educate users on photographic nuances.