Adobe has updated its experimental Indigo camera app to version 1.1, adding a set of AI tools called "AI Playground." In a notable choice, these tools do not run on Adobe's own AI models (called Firefly). Instead, they are powered by Google's Nano Banana AI model, according to a blog post by Adobe fellow Marc Levoy on Adobe Research's website (The Verge).

Using a rival company's AI model instead of its own is a deliberate choice, at least for now. Adobe says it may switch away from Nano Banana or add support for other models in the future. The company describes AI Playground as an experiment, not a finished product.

AI Playground is split into four sections: Styles, Object Editing, Photo Guidance, and Custom Edit. Styles changes the overall look of an image using AI. Object Editing lets users change or remove things in a photo. Photo Guidance gives advice on how to take a better shot. Custom Edit is the most open-ended, letting users make creative AI changes to their photos. Adobe did not go into detail about how each section works.

Free access to all of AI Playground, with no account needed, is being tested with a small percentage of Indigo users over the coming weeks. Levoy said Adobe may eventually offer a paid version if the feature proves popular, which suggests the company is testing whether people actually want this before committing to it.

Project Indigo launched as a completely free app with no account requirement, from Adobe Labs, the company's research division. It was built by a team called Nextcam (Gigazine; Adobe Research). The app is a computational photography camera app, which means it uses software to enhance and manipulate photos as you take them, rather than relying only on the camera lens. Contributors Jiawen (Kevin) Chen and Zhoutong Zhang are credited on the project's Adobe Research page. The original release was meant to explore what computational photography can do when freed from the pressures of a commercial product schedule. Adding generative AI extends that exploration into new territory.

The model choice is what may catch the attention of people who follow the tech industry. Adobe has invested heavily in Firefly as its own AI system, designed to be safe for commercial use and integrated across its Creative Cloud apps. Routing an Adobe-branded feature through Google's model instead raises questions about speed, whether the processing happens on the phone or in the cloud, what the licensing terms are, and where Adobe sees the limits of its own AI. The announcement does not answer these questions, but Adobe's willingness to swap or add models suggests it sees the AI model as a replaceable part rather than a permanent decision.

There is a broader trend worth understanding here. Over the past several years, smartphone cameras have increasingly used software to do things that used to require editing a photo after taking it — combining multiple shots into one, identifying objects in a scene to adjust them separately, and adjusting lighting in real time. Generative AI in the camera app itself, applied right when you take the photo, is a natural next step in that direction. What is less clear is whether people actually want AI manipulation built into the act of taking a photo, or whether they prefer it as a separate step afterward. Adobe's experiment is partly a test of where that line should sit.

The free, no-account approach also matters. Adobe's main software products have long required subscriptions and account sign-in. By making Indigo and its AI tools free with no sign-on, Adobe is keeping this in the experimental category, where it can see how much engagement the feature gets before deciding whether to turn it into a real product. Levoy's mention of a possible paid version is the clearest sign that Adobe could commercialize this if enough people use it.

For developers and AI professionals, the key open questions are which AI model Adobe will settle on, whether the processing happens on the device or in the cloud, and whether the AI features work on photos after they are taken, on a live preview, or both. The four-section structure suggests a range of AI capabilities in one interface, but how well each one works will become clearer as the test reaches more users.

Adobe has not given a timeline for broader availability or for the possible paid version. The experiment is live now for the small percentage of Indigo users included in the initial test.