Vue normale

Adobe wants you to play with generative AI in its Project Indigo camera app

The experimental tools enable distraction removal, art style–based edits and more, but only for a small subset of users.

A side-by-side image of the same ornate metal lantern, shown on a stone pedestal in a lush garden on the left and on a rocky mountain ledge with dramatic peaks at sunset on the right.Adobe

Last year, Adobe’s Nextcam team launched Project Indigo, an experimental camera app for iPhones that offered full manual controls and promised a more natural look from smartphone photos. Now, the team has announced AI Playground, an experimental generative AI tool inside of Project Indigo.

In its announcement, Adobe explains that text-based editing (e.g., asking an AI chatbot to make changes to a photo for you) is still clunky, both in its results and usability. This new tool aims to improve that experience by keeping more of the process – both capture and edit – in one app. It says that AI Playground aims to “explore how generative AI can assist mobile photographers at the point of capture.”

AI Playground offers a handful of buttons that, when tapped, trigger specific generative AI prompts engineered by Adobe. These include options for removing distractions, creating a shallow depth of field, applying certain styles, changing lighting and critiquing photos. There’s also a Custom Edit button where users can input their own prompts, either via typing or dictation. For now, Adobe says it is using Google’s Nano Banana model, but that is subject to change, and may even change on a per-button basis.

A photo of a stone garden lantern with a mobile AI editing interface overlaid in the center, flanked by the original garden scene on the left and a hand-drawn-style illustration of the same lantern on the right.
For now, the AI Playground offers four different buttons: Object Editing, Styles, Photo Guidance and Custom Edit.
Adobe

The “Styles” button allows users to apply a range of preset styles, turning photos into drawings or paintings or changing the lighting. Adobe says the available options “represent techniques we have found to be most successful when applied to everyday snapshots,” so it isn’t an endless list. It also clarifies that it trained the model to identify and highlight the main subject to keep focus on the most important parts of the photo.

A side-by-side image of a smiling woman in an art gallery, first with a busy crowd behind her on the left and then with the background cleared of people so only the empty gallery and paintings remain on the right.
The AI Playground allows users to remove distractions in the camera app, instead of opening a separate editing app.
Adobe

Object Editing offers tools familiar to editing apps, but Adobe says it is “exploring whether simple editing operations can be performed in the camera app using a generative model.” There are buttons for removing distractions and creating a shallow depth of field. If those specific tools don’t cover what you are looking to do, the Custom Edits button lets you change almost anything, similar to standalone generative AI tools. That includes completely swapping out backgrounds and specifying different styles. Adobe also showed off the ability to turn a drawing into a photo-like image, or change a sculpture into a “real” person.

A side-by-side image in an art gallery showing a white marble statue of a woman leaning to listen on the left, and on the right the same pose transformed into a lifelike young woman in flowing white clothing standing on the statue’s base.
The AI Playground lets you completely reimagine photos, including turning sculptures or paintings into photo-realistic-style people or scenes.
Adobe

Adobe also highlighted the feedback section, which looks to be a more robust version of Google’s “Camera Coach,” albeit one that critiques photos after the fact, rather than giving you suggestions as you’re composing. It says that it is aimed at helping users learn or advance their photography by providing critique. The tool can provide both reshooting suggestions since it operates within a camera app, along with editing suggestions. Adobe says it is “working on customizing these prompts to reflect individual phone models and lenses, each of which has unique capabilities,” so that users get suggestions specific to what their phone is capable of.

There are some caveats with the AI Playground. First, while it’s free (for now) and doesn’t require sign-in, Adobe is only rolling this out to a small percentage of Project Indigo users. It says that if AI Playground is successful, there will be a paid version. Adobe also makes it clear that this is an experiment inside of an experimental app, so things are subject to change. You need to be connected to the internet to use AI Playground, and to save transmission and generation time, images are 2K pixels on a side.

A three-panel image showing a man in glasses lit dramatically at a table on the left, and on the center and right his photo displayed within a mobile “AI Playground” interface that provides detailed photo critique and reshooting tips.
The critique tool can provide detailed feedback on how to improve an image.
Adobe

The Nextcam team also explains that it has had to create a workaround for HDR images, as image-generation models only output SDR images at this point. So, if you use Nano Banana to edit an HDR photo (even one taken with Project Indigo), it will produce an SDR image. To address this, Adobe says it built a custom ML model to post-process the SDR image and boost it back to HDR, but warns that it isn’t always accurate.

The announcement also briefly addresses ethics as well. Adobe says it is “working to attach metadata to AI-edited images produced from the AI Playground by using technology from Adobe’s Content Authenticity Initiative,” and specifies that some models, such as Nano Banana, apply invisible watermarks to images. At the end of the day, though, Adobe puts the onus on users: “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.”

A cyclist in a bright jacket rides up a rural road lined with grass and trees, shown first in heavy gray fog and then in warm golden sunlight, highlighting a dramatic change in lighting.
The AI Playground means you can completely alter a photo with a simple prompt.
Adobe

Ethics aside, it seems like an interesting move for the app, given that the original goal of Project Indigo was to produce more natural images that offer “SLR-like” quality from your iPhone. DPReview’s Dale Baskin even chose Project Indigo as his Gear of the Year last year, because it unlocked the ability to take more natural images, encouraging him to use his phone for photos more often. The AI Playground tools, however, seem to take things in a very different, more synthetic direction.

Additionally, the results, at least for now, don’t necessarily inspire confidence. The image at the top of this article looks more like a video game rendering than a photograph, with a distinctly synthetic sheen rather than the ‘natural’ look Indigo set out to offer. Plus, in some instances, such as the cycling image above, Adobe acknowledges that the models will change details that you don’t want changed. That includes altering face shape and expressions, even if it isn’t part of the user’s prompt or intent. The company’s workaround is to upload reference photos, which certainly isn’t ideal.

For those interested in testing AI Playground, you’ll need to update the Project Indigo app to version 1.1. If you are one of the selected users, you’ll see a screen pop up inviting you to use the feature. If you’re not a fan of GenAI in your camera app and you get invited, you can simply decline and keep using Project Indigo as you did before. Adobe also says that the Nextcam team is working on non-GenAI computational photography features that it plans to launch soon.

About the Author: Abby Ferguson holds an MFA in photography and blends a fine-art background with extensive editorial and commercial experience. She built and led an in-house imaging program for a large vacation rental company and has served as an associate editor at Popular Photography. She also teaches at a local community college and serves on its Photographic Technology Program Advisory Board, sharing her knowledge and enthusiasm for the medium.

Want to stay up to date on the latest product news and releases? Add DPReview as a preferred source to ensure our independent journalism makes it to the top of your Google search results.

add as a preferred source on google

An RX100 is surely next, why can’t I get excited about that?

The RX100 series represented a big breakthrough for image quality, back in 2012, but the six iterations since then suggest it’s not about to turn into my dream camera.

Sony RX100 V backlit in front of out-of-focus autumnal treesCarey Rose

I’m not completely shocked that Sony updated the RX10: the existing model was clearly out of production and increasingly silly second-hand prices suggested there was pent-up demand. And now the work of integrating more modern components with a Type 1 sensor has been done, surely an updated RX100 will be next?

I found myself thinking about what that might look like, and the more I do, the more I suspect it won’t be what I want.

Two years ago I made the argument that there is a market for enthusiast compacts: small cameras with decent image quality and sufficient direct control to give a satisfying photographic experience. I proposed an RX100-style camera with a less-expensive sensor and the original 24-70mm equiv F1.8-2.8 lens. Sadly, that’s not what I’m expecting Sony to make, at all.

Sony is a technology company. A technology company that makes some very good cameras, but ultimately a company that believes that their cameras will always be made better by throwing more and newer technology into them.

The RX100 series have always been products whose engineering I admired, rather than cameras I enjoyed

So instead of the stripped-back camera with direct controls and a (relatively) accessible price tag that I’d love someone to make, I suspect we’ll get a camera with, at the very least, all the bells and whistles of the RX10 V. I’m going to guess it’ll have the rangier but slower 24-200mm equiv lens, the ability to shoot at framerates around double the speed of a pro-targeted DSLR and a list price of at least the $1500 that the VII’s launch price would equate to, today.

And thinking about this made me remember how frustrated I’ve always been with the RX100 cameras. Because, with their introduction of the Type 1 (13.2 x 8.8) sensor to the compact market, they represented a huge step forward in terms of how much image quality you could fit in your pocket. The JPEG color on the first one was perhaps a little wonky but that 116mm2 sensor was nearly three times the size of the Type 1/1.7 (43mm2) chip that had dominated the space up until then, and that’s the kind of difference you could see in every photo.

Top-down view of an Olympus XZ-2 in someone's hand
The Olympus XZ-2 iHS had a relatively small sensor but a bright lens with a useful range. It also had a clicking dial around the lens that gave good feedback when changing settings and worked well with the touchscreen interface. It also had a switch that de-clicked the ring and engaged a different function, for additional control.
Richard Butler

But despite the appearance, price and that short, bright lens, the RX100 never felt like an enthusiast compact. Compared to the likes of the Olympus XZ models or Canon’s G and S-series cameras, the Sony sorely lacked engaging control points. Typically I’ll use an enthusiast compact in Aperture Priority mode, so would want to be able to control the aperture value and exposure compensation: both of them settings that adjust in fixed, stepped intervals. The RX100 doesn’t have a well-placed clicking control point, meaning you have to change most settings via the vague, free-rotating ring around the lens. There’s a lag before it recognizes that your input is intentional and there’s no tactile feedback that you’ve changed anything.

The result was a user experience that left you feeling detached from the photo taking process

The result, even before you visited the old, rather chaotic menus, was a user experience that left you feeling detached from the photo taking process: the precise opposite of what I believe a standalone camera needs to offer. The RX100 series have always been products whose engineering I admired, rather than cameras I enjoyed.

I remember being advised “stop trying to make it do what you want, and just use it as a point-and-shoot,” and finding that helped. But if I wanted to just take snaps, I’d probably use a smartphone, because the phone makers know their devices aren’t about taking control, so they’re happy to shoot and combine multiple images with every shot, letting them rival or exceed the image quality of a Type 1 camera taking single photos.

I could be wrong, of course. Sony might significantly rework the RX100 to make it fun to take control of. Or it might not make one at all. And, even if it does, I’m not going to assume that the things I, personally, want are the things that the market as a whole wants: Sony will have a much better idea of what sells and what doesn’t, and of what would-be customers are asking for.

But, while the whole thing remains hypothetical, you can’t blame me for hoping.

About the Author: Richard graduated as a scientist but had a lot more fun writing and shooting for his university magazine. A number of years spent variously as a reporter, writer and editor on science and engineering titles combined his knowledge of science with his interest in images and words. But it was spotting the connections between emission spectra, white balance and all the nonsense he’d taught himself playing around in Photoshop that helped kindle an interest in digital photography. Searching for a camera led to him discovering DPReview and Richard was recruited by Phil Askey in 2007. He’s been combining his love of photography, communication and attention to detail (pedantry?) ever since. He has unusually strong opinions about lenses for the APS-C format.

Want to stay up to date on the latest product news and releases? Add DPReview as a preferred source to ensure our independent journalism makes it to the top of your Google search results.

add as a preferred source on google

❌