Adobe has updated its experimental Indigo camera app to version 1.1, adding a suite of generative AI tools called "AI Playground" that, notably, does not run on the company's own Firefly models. Instead, the features are powered by Google's Nano Banana AI model, Adobe fellow Marc Levoy disclosed in a blog post on Adobe Research's website (The Verge).

The decision to use a third-party model rather than Firefly is a deliberate one, at least for now. Adobe says it may switch away from Nano Banana or introduce additional model support over time. The framing throughout is explicitly experimental; Adobe describes the new AI Playground tooling as an experiment rather than a finished product feature.

AI Playground is organized into four sections: Styles, Object Editing, Photo Guidance, and Custom Edit. The Styles section applies generative transformations to the visual character of an image. Object Editing allows users to modify or remove elements within a frame. Photo Guidance offers direction on how to improve a shot, and Custom Edit provides a more open-ended path for generative manipulation. The specific mechanics of each section were not detailed beyond these categories.

Free access to the full AI Playground suite, with no sign-on requirement, is being tested with a small percentage of Indigo users over the coming weeks. Levoy said Adobe will eventually offer a paid version if the feature proves popular, which signals the company is treating this as a demand-validation exercise rather than a commitment to a permanent free tier.

Project Indigo itself launched as a completely free, no-account-required app from Adobe Labs, the company's research division, developed by its Nextcam team (Gigazine; Adobe Research). It is a computational photography camera app, and contributors Jiawen (Kevin) Chen and Zhoutong Zhang are credited on the project's Adobe Research page. The app's original release positioned it as an exploration of what computational photography can do when freed from the constraints of a shipping product roadmap. Adding generative AI to a camera app that already emphasizes computational image processing extends that exploration into the generative space.

The model choice is the element most likely to catch the attention of industry watchers. Adobe has invested heavily in Firefly as its brand-safe, commercially licensed generative AI offering across Creative Cloud applications. Routing an Adobe-branded generative feature through Google's Nano Banana model rather than Firefly raises practical questions about latency, on-device versus cloud inference, licensing terms, and where Adobe sees the boundaries of its own model ecosystem. These questions are not answered in the current announcement, but the explicit openness to swapping or adding models suggests Adobe is treating the underlying model as a pluggable component rather than a fixed architectural decision.

Worth flagging is the broader pattern this fits into. Computational photography has been steadily absorbing techniques that were once the domain of post-capture editing: multi-frame fusion, semantic segmentation, and real-time tone mapping are now table stakes in smartphone camera pipelines. Generative AI in the camera app itself, applied at or near the point of capture, is a natural extension of that trajectory. What is less settled is whether users want generative manipulation as part of the capture flow or as a distinct post-processing step. Adobe's experiment is, among other things, a test of where in the workflow that boundary sits.

The no-sign-on, free-access framing also matters. Adobe's commercial software has historically been gated behind Creative Cloud subscriptions and account authentication. Indigo's no-account requirement and the AI Playground's free, no-sign-on test period place this firmly in the labs-and-experiments category, where Adobe can gauge engagement and gather usage data before committing to productization. Levoy's mention of a potential paid version is the clearest signal that commercialization is on the table if uptake justifies it.

For developers and AI practitioners, the details that matter are the model orchestration question (Nano Banana today, potentially others tomorrow), the on-device versus cloud inference split (not specified in the announcement), and whether the generative features operate on captured stills, live previews, or both. The four-section structure suggests a range of generative capabilities bundled into a single interface, but the depth and quality of each remain to be assessed as the rollout expands beyond its initial small test cohort.

Adobe has not committed to a timeline for broader availability or for the potential paid tier. The experiment is live now for the limited percentage of Indigo users included in the initial test.