Adobe last week began pushing an update to its Project Indigo experimental iPhone camera app that grafts a generative AI playground onto the moment of capture—and it’s not using its own Firefly models to power the magic. The new AI Playground, built with Google’s Nano Banana image-editing models, can delete strangers from beach shots, fake a shallow depth of field, turn a snapshot into a pen-and-ink illustration, or tell you how to reshoot the scene before you walk away. The feature arrived in version 1.1 as an opt-in test for a small, randomly chosen slice of users, free and with no Adobe account required, but the company warns it won’t stay free if it catches on.

What’s New in Indigo’s AI Playground

AI Playground lives directly in Indigo’s filmstrip—tap the sparkle icon beneath a photo you just took and you enter a workspace with four sections. Object Editing removes background people, vehicles, signs, and clutter, then fills the holes with generated pixels; it also simulates portrait-mode background blur. Styles reimagine the photo as an ink sketch, watercolor, or apply golden-hour lighting. Custom Edit lets you type or dictate free-form instructions (“remove the fence in the foreground and the people in the background”), save prompts, and rerun them for variations. None of these operations are deterministic—tapping the same button again typically produces a slightly different result.

Photo Guidance is the most unusual tool. Instead of fixing a finished image, the AI analyzes your shot and suggests what to change while you can still reshoot: move closer, switch lenses, wait for the sun to shift. Adobe says the advice is being tuned to account for your specific iPhone model’s lenses and limitations, but early testers should treat suggestions as starting points rather than gospel.

Because the generative heavy lifting happens in the cloud, editing speed depends on your connection, and Adobe is transparent that today’s free experiment implies a future bill. “If the feature proves popular, we will eventually offer a paid version,” Adobe fellow Marc Levoy wrote in the announcement. The company has not yet disclosed pricing, credits, or subscription models.

Why Google Instead of Firefly?

Adobe’s decision to partner with Google for the AI model surprised many observers who associate the company’s generative ambitions with Firefly—the in-house system trained on licensed content and tightly integrated with Photoshop, Lightroom, and other Creative Cloud apps. Levoy described the choice as pragmatic: “We’re using Google’s Nano Banana to start, but we may switch or introduce additional model support over time.”

The move fits a pattern emerging across Adobe’s product line. Rather than force every AI feature through a single model, the company increasingly acts as an orchestrator—picking the best engine for the job while controlling the user experience, metadata pipeline, and creative workflow. For Indigo, that means Adobe’s Nextcam team focuses on capture processing, interface design, and the bridge to Lightroom, while Google supplies the raw image-synthesis capability. This is less a rejection of Firefly than a recognition that no one model leads in every category, and that a camera app with zero sign-in friction needs a plug-and-play AI service.

The downside is dependence. Google could change pricing, restrict APIs, or shift the model’s safety policy, leaving Adobe scrambling. Creative pros who pick Adobe explicitly for Firefly’s indemnification and content-credential promises may wonder why an Adobe-branded camera is depending on a third-party model that applies Google’s invisible SynthID watermark—but not yet the full C2PA provenance chain Adobe champions.

From Capture to Cloud: What It Means for Your Workflow

Indigo has always been an iPhone-first experiment, but its output doesn’t stay on the phone. The app can send JPEG or DNG raw files directly into Lightroom mobile, preserving embedded looks and HDR data. From there, photos sync to Lightroom desktop and Photoshop on Windows—meaning a generative edit applied seconds after capture in Indigo can land on your PC minutes later.

For Windows users who manage photo libraries on their desktop, this creates a new responsibility: verify what came from the sensor and what came from a model. A background erased with Google’s AI might look flawless at phone-screen size, but zoom in on a 4K monitor and you could spot identity drift in faces, garbled text on storefronts, or reflections that don’t match the scene. Adobe is working on attaching C2PA credentials to edited images, but that metadata is fragile; it can be stripped by social media uploads, screenshots, or simple file conversion.

In practical terms, treat an AI-edited Indigo shot like a baked JPEG—keep the original raw file untouched. If you later need to prove what was really in front of the lens, you’ll want the unaltered sensor data.

The Authenticity Question: Keeping Your Edits Honest

Object removal and restyling sit on a spectrum that runs from trivial cleanup to documentary deception. Removing a soda can from a vacation photo is one thing; erasing a person from a protest scene is another. Adobe acknowledges the tension. The company says it will add C2PA Content Credentials “in a future update” and notes that Google’s SynthID watermark already tags outputs from Nano Banana. But that watermark is invisible and meant for detection by automated systems, not casual viewers.

On a Windows PC, you may not see any indication that an image was altered unless the app that receives it—File Explorer, Photos, a web browser—knows to check for credentials. Microsoft has been adding support for C2PA in some tools, but adoption is not universal. Until provenance becomes a standard part of the operating system, the burden falls on the photographer to label files and the viewer to ask questions.

Adobe’s Marc Levoy summed up the philosophical stance: “Whether the edits you apply are ‘fair’ or ‘misleading’ depend on what you do, who you share the images with, and how you present them.” In other words, the tool won’t police intent. That puts a premium on your own documentation habits.

Should You Use It? A Practical Guide

If you’re among the testers selected for AI Playground, the payoff can be immediate. A quick background cleanup can rescue a shot that would otherwise be deleted, and Photo Guidance might genuinely teach you to frame better—provided its advice stays realistic for your iPhone model.

But set a few personal rules:
- Always preserve the original. Indigo lets you keep unedited versions; use that feature. Rename or star the original in Lightroom so you don’t accidentally replace it.
- Don’t rely on AI edits for anything you might need to verify later. That includes insurance claims, legal evidence, or portfolio work that demands authenticity.
- Check your export settings. When sending a photo to Windows, verify whether the metadata (including any C2PA manifest) survives the trip. If you use cloud sync, test with one image first.
- Limit experimentation to Wi-Fi. Generative edits eat data and battery; long waits on cellular can kill the spontaneity Indigo otherwise encourages.
- Take Photo Guidance with a grain of salt. If the assistant tells you to “use a faster lens” or “widen the aperture” on a fixed-aperture iPhone, you’ll know the device awareness still needs polishing.

For those not in the test group, there’s little to do right now other than update Indigo to version 1.1 so you’re ready if access expands. Adobe has not published a roadmap or waitlist; your best bet is to watch the Adobe Labs blog and the app’s settings for a toggle to appear.

What’s Next for Adobe Indigo and AI Photography

The immediate milestone is whether Adobe widens the test pool or extends the free period beyond the initial “few weeks.” Usage patterns matter—if people mainly play with Styles and ignore Photo Guidance, the project may tilt toward entertainment rather than education. Conversely, if guidance proves popular, Adobe could weave it into Lightroom’s mobile camera or even the desktop develop module.

Pricing will be the next flashpoint. Adobe has experience with generative credits in Photoshop and may bring a similar model to Indigo: a monthly bucket of AI edits, with top-ups available. A subscription tied to Creative Cloud would appeal to existing users, while a standalone in-app purchase could draw new customers—but both require account login, which breaks the current anonymous design.

Android support, promised at Indigo’s launch last year, remains an open item. An Android version would not only double the possible test audience but also deepen Adobe’s collaboration with Google, since Android’s camera APIs and chipset diversity would stress-test the device-awareness claims.

Finally, the model strategy will evolve. Adobe says it may add Firefly or other engines down the road. That would allow a tiered approach: Firefly for commercial-safe edits with full indemnification, Google for broader creative restyling. The interface could remain the same—tap “remove clutter” and let Adobe pick the model behind the scenes—but professionals will demand clear on-screen labeling and durable metadata that travels with the file to Windows.

Project Indigo’s experiment with Google AI is a reminder that the most consequential camera updates aren’t always about megapixels or lenses. Sometimes they’re about what happens in the seconds after you press the shutter, and whether the software that transforms your photo also preserves your trust.