Photoshop's NEW On-Device AI Remove Tool

Adobe shipped Photoshop 27.7 on the 19thMay 2026, and tucked inside it is one of the more genuinely useful quality-of-life features I've seen in years: the ability to run the Remove tool's generative AI model entirely on your own hardware.

For photographers and retouchers, that means object removal that's faster, works offline, and keeps sensitive client work on your own machine.

What the Remove Tool Actually Does

The Remove tool sits alongside the Healing and Clone tools as Photoshop's go-to way of erasing distractions. You paint over something you don't want, a stray tourist, a power cable, a logo, a blemish, and Photoshop fills the area with believable pixels sampled and synthesised from the surroundings.

In the current version there are effectively three flavours of Remove.

Standard Remove (Generative AI off)

Uses content-aware style algorithms, processed locally. It works well on simple, repeating backgrounds, but it can struggle with complex textures and edges.

Cloud Generative Remove (Generative AI on, Cloud)

Sends your selection to Adobe's servers, where a Firefly-class model generates new pixels to fill the gap. It generally performs better on complex scenes with fine detail, like foliage, hair and signage.

On-device Generative Remove (Generative AI on, Device)

New in Photoshop 27.7. A generative model is downloaded to your computer so the Remove tool can do that heavy lifting locally instead of relying on the cloud.

Functionally, on-device mode aims to give you cloud-quality results with the reliability of local processing.

Why Bother With On-Device Remove?

If Remove already runs in the cloud, why bother with on-device at all? Three reasons.

Speed and responsiveness

In on-device mode, the model runs directly on your GPU rather than making a round trip to Adobe's servers. On powerful machines that means noticeably snappier results, especially on large layered documents or when you're doing lots of small removals across a full shoot.

It works when you're offline

Cloud Remove simply doesn't function without an internet connection. If you're editing on a train, in a remote landscape, in a studio with poor Wi-Fi or at a client office behind a locked-down network, that's a problem. On-device Remove works exactly the same way whether you're online or offline.

Better privacy and data control

Some jobs involve embargoed campaigns, confidential documents or sensitive subjects. In those situations it's often preferable, or contractually required, to keep all processing on your own hardware. On-device Remove keeps the pixels on your machine, which can be an important reassurance for privacy-conscious clients.

The Catch: The Hardware Bar Is High

Running a generative model locally is demanding. Photoshop checks your system automatically, and if it doesn't qualify, the Device option in the Remove tool is simply greyed out.

On Windows, Adobe's current guidance calls for a modern multi-core CPU, a strong GPU, and decent RAM and SSD space:

On top of that, Adobe lists higher VRAM thresholds by GPU family. Current documentation and support posts indicate that many mid-range GPUs, some RTX 30-series cards with 12GB of VRAM for example, still don't unlock the Device option for Remove, even though they run other AI features comfortably.

On Mac, on-device Remove is Apple Silicon only. You'll need an M1 Pro or later (M1 Max, M2 Pro, M2 Max and newer are recommended for smooth use), 24GB of RAM or more, and macOS Tahoe (26.4) or newer. Intel Macs aren't supported for the on-device model.

How to Use the On-Device Remove Tool

Once your hardware qualifies and you're running Photoshop 27.7, it's all controlled from within the Remove tool itself.

1. Select the Remove tool from the toolbar, where it sits with the Spot Healing Brush and Healing Brush.

2. Open the Mode dropdown in the Options bar, where you'll see options such as Auto, GenAI On and GenAI Off.

3. Choose Device for Generative AI processing. Within that panel, find the Generative AI processing setting and switch it from Cloud to Device.

4. Download the on-device model. The first time you select Device, Photoshop prompts you to download it, just under 5GB. Confirm it and let it finish.

5. Paint away distractions. With Device selected, paint over unwanted objects as usual, and Photoshop will use the local model rather than the cloud to generate the replacement pixels.

6. Switch between Cloud and Device whenever you like, from the same Mode dropdown, if you want to compare quality or fall back to the cloud on less powerful hardware.

If the Device option stays dimmed, it usually means either your hardware doesn't meet the minimum spec or GPU acceleration is switched off in Photoshop's Performance preferences.

Who Will Benefit Most?

On-device Remove is particularly valuable for three groups.

Location, travel and event photographers

You often edit in the field or away from reliable internet. On-device Remove keeps your clean-up workflow working exactly the same wherever you are, hotel room, train, remote shoot or studio with spotty Wi-Fi.

High-volume retouchers

If a big chunk of your day goes on removing logos, stray hairs, dust, cables and background clutter, a faster, offline, privacy-friendly Remove tool makes that daily work smoother and more predictable, especially on time-critical jobs.

Commercial and editorial shooters with sensitive work

When you're dealing with unreleased campaigns or sensitive subjects, being able to say all the retouching was done on-device is a genuine advantage with clients and legal teams.

For casual users on modest hardware, the existing standard and cloud Remove modes will carry on doing the job perfectly well. But if your machine meets Adobe's on-device requirements, the new Remove model in Photoshop 27.7 is one of the most meaningful quality-of-life upgrades in the current release, especially if you live in the Remove tool all day and want its full power without always depending on the cloud.

Instantly Fix "Impossible" Glasses Reflections in Photoshop

Removing reflections from glasses has always been one of those jobs in Photoshop that's either felt impossible or just painfully tedious. In this tutorial, I'm showing you how the new Firefly Image Model 5 in the Photoshop Beta handles this specific problem in a way that I think you're going to find really useful.

I'm working with a portrait of Thomas Coulter, one of the veterans from my 39-45 Portraits Project, to walk you through exactly how it works.

The Challenge with Older Models

If you've tried using Generative Fill for this before, you'll know that older models like Firefly Image 1 could certainly remove a reflection, but they often introduced other problems at the same time. You'd sometimes end up with subtle changes to facial structure, eyebrows, or the shape of the glasses frames themselves. The reflection might be gone, but the portrait no longer looked quite right.

Why Firefly Image Model 5 is Worth Knowing About

Model 5 has been built with detail preservation as a priority. The idea is that it only changes what you've asked it to change, leaving everything else as close to the original as possible.

Worth knowing: this is a premium model, so it uses 10 generative credits rather than one. It also only produces a single variation, but given the quality of the result, that's rarely a problem.

How to Do It, Step by Step

1. Open Photoshop Beta - You'll need the Beta version to access the latest Firefly models. [00:56]

2. Make your selection - Use the Selection Brush Tool to paint over the reflections on the lenses. You don't need to be overly precise; going slightly over the frames is fine. [01:25]

3. Open Generative Fill - Click Generative Fill in the Contextual Taskbar. If you can't see it, go to Window > Contextual Taskbar. [01:49]

4. Choose the right model - This is the key step. In the Taskbar settings, look under Adobe Models and select Firefly Image Model 5 (Preview). [06:16]

5. Enter your prompt - Something simple like "remove the reflection from the glasses" is all you need. [05:32]

6. Generate - Hit Generate and give it around 10 to 12 seconds. [06:21]

The Results

What I find genuinely impressive here is that once the reflection is gone, everything else stays exactly as it was. The eyebrow hairs, the skin texture, the precise shape of the frames - all identical to the original file.

Now, Camera Raw and Lightroom do have reflection removal tools built in, and they're well worth trying, particularly on larger reflections. But for detailed areas like eyewear, where precision really matters, this approach in Photoshop gives you a level of control and accuracy that's hard to beat. If you've got portraits sitting in your archive that you've written off as too difficult, this is a good reason to dig them back out.

Fix IMPOSSIBLE Backgrounds Instantly ( Lightroom + Photoshop )

Recently, Steven Gotz, a member of the Photography Community on SKOOL ( LINK ), sent over a brilliant RAW file of a condor. Stunning subject, great light, one problem: a massive fence running right through the background.

Rather than leave it on the shelf, I figured it was the perfect excuse to put the latest updates in Lightroom and Photoshop Beta through their paces. What would have taken ages with the Clone Stamp tool a couple of years ago can now be sorted in seconds. Here's exactly how I did it, using two different workflows.

Workflow 1: Photoshop Beta with Firefly Image 5

This is the quickest route right now, and the results are genuinely impressive.

The key is using the new Firefly Image 5 (Preview) model inside Photoshop Beta. It's been built specifically for editing while preserving detail, which matters a lot when you're dealing with complex textures like feathers and rocky backgrounds.

1. From Lightroom to Photoshop Beta. Right-click the image in Lightroom and choose Edit In > Adobe Photoshop Beta.

2. Select All. Once you're in Photoshop, go to Select > All. This gives the AI the full context of the frame before you do anything.

3. Switch to Firefly Image 5. Click Generative Fill in the contextual taskbar. Here's the bit that matters: don't use the standard model. Switch it to Firefly Image 5 (Preview) from the dropdown.

4. The prompt. This model needs a prompt to work, unlike some of the others. I kept it simple: "remove the fence from this picture."

5. Refine the detail. The AI did a great job on the background, but because Firefly Image 5 currently outputs at 2K, the fine detail around the bird's eye and feathers was slightly softer than the original RAW. The fix is straightforward: use the Object Selection Tool to select the bird and the rock, then fill that area on the layer mask with black. That reveals the sharp original bird while keeping the AI-cleared background intact.

Workflow 2: Lightroom to Firefly Web

Not on the Photoshop Beta? No problem. You can get to the same place via Lightroom's sharing feature.

1. Share to Firefly. In Lightroom, hit the Share button (top right) and select Firefly: Edit an image. This opens your browser and drops the photo straight into the Adobe Firefly web interface.

2. Settings and generate. Select Firefly Image 5, bump the resolution to 2K, use the same prompt ("remove the fence from this picture"), and hit generate.

3. Back to Photoshop. Download the cleaned image, go back to Lightroom, and open the original file in the regular version of Photoshop.

4. Stack and align. Use File > Place Embedded to bring the Firefly-cleaned version in on top of your original. Rasterise the top layer, select both layers, then go to Edit > Auto-Align Layers to make sure everything lines up perfectly.

5. The masking trick. Same principle as Workflow 1: use the Object Selection Tool to select the bird and the rock, then hold Option (Mac) or Alt (Windows) and click the mask icon. This hides the AI version of the bird and brings back the sharp, high-detail original underneath.

Why the masking step matters

This is the part I think is really important. It's not about letting AI take over the whole image. It's about using it to fix a specific problem, in this case the background, while keeping the actual subject exactly as it was captured in the RAW file. The integrity of the original is what you're protecting.

Have a look through your archives. Chances are there are shots you wrote off because of something in the background. It might be worth giving them another look.

NEW 💥 Photoshop's One-Click Auto Distraction Removal

Adobe has just dropped a seriously powerful update to the Remove Tool in the Photoshop Public Beta (version 27.6.0), and it’s a total game-changer for cleaning up your photos. It can now automatically scan your entire image, identify distractions across 26 different categories, and let you remove them with a single click.

Here is a quick look at how it works and how you can start using it to save yourself hours of manual cloning and healing.

What is the New "General Distractions" Feature?

Previously, the Remove Tool had specific buttons for "Wires and Cables" or "People." This new update introduces General Distractions. It uses generative AI to find things like trash cans, signs, vehicles, and even stray animals that might be cluttering up your shot.

How to Use It: A 3-Step Tutorial

Before you start, make sure you have GPU hardware acceleration turned on in your Photoshop settings (Preferences > Performance) to ensure the tool runs smoothly.

1. Select the Remove Tool

Head over to your toolbar and select the Remove Tool. In the options bar at the top, make sure Sample All Layers is ticked and, most importantly, check the Create New Layer box. This acts as a fail-safe, putting all your removals on a separate layer so you can easily bring things back if you change your mind later.

2. Find Your Distractions

In the options bar, click on the Find Distractions dropdown and choose General Distractions, then click Find. Photoshop will take a few moments to scan the image. When it’s finished, it will highlight potential distractions with color-coded overlays.

The cool part? The list of categories it shows you is dynamic. It won't show you all 26 categories; it only lists the ones it actually found in your specific photo—like "Vehicles," "Animals," or "Urban Elements."

3. Refine and Remove

You have total control over what stays and what goes:

• Toggle Categories: You can untick specific categories in the dropdown if Photoshop picked up something you actually want to keep (like a cool cloud it mistook for a "light diffusing element").

• Manual Overwrite: Use the plus (+) or minus (-) brush icons in the options bar to manually add areas to be removed or protect areas you want to save.

• The Big Reveal: Once you're happy with the selection, click the Tick icon. Photoshop will work its magic, and the distractions will vanish, seamlessly filling in the background.

Why This Matters

I've been testing this on complex street scenes and busy beach shots, and the results are mind-blowing. It handles everything from removing pigeons at someone's feet to rebuilding stone walls where a trash can used to be. It’s not just a time-saver; it’s doing work that used to require advanced cloning skills in just a few seconds.

Since this is currently in the Public Beta, if you run into anything unexpected, be sure to use the "Feedback" icon in the top right of Photoshop to let Adobe know. The more feedback we give them now, the better the final version will be.

The Photoshop Zoom Setting You NEED to Change ✅

Whether you are just starting out with Photoshop or you have been using it for years, there is one specific setting that can occasionally make it feel like the software is behaving rather strangely. I wanted to share a quick tip about the Zoom tool that might just save you a bit of frustration.

The Mystery of the Shifting Zoom

Have you ever tried to zoom in on a specific detail, only for that area to suddenly jump to the middle of your screen? Usually, when you click with the Zoom tool, you expect the image to get larger exactly where your cursor is sitting. However, there is a setting that changes this behaviour entirely.

If your image keeps repositioning itself every time you click to magnify, it is likely because of a single tick box in your preferences.

How to Fix It

Depending on whether you are using a Mac or Windows, the menu location is slightly different, but the setting itself is the same:

• On Mac: Go to the Photoshop menu, then Settings, and select Tools.

• On Windows: Go to the Edit menu, then Preferences, and select Tools.

Look for the option labelled Zoom Clicked Point to Centre.

If this is ticked, Photoshop will take the exact point you clicked and move it to the very centre of your workspace as it zooms in. If you find this distracting, simply uncheck the box. Once you do that, your zoom will behave in the traditional way, staying put exactly where you click.

Why Would You Use It?

You might wonder why this setting even exists if it feels so counter-intuitive at first. It actually comes in quite handy when you are working on very large, high-resolution images or wide landscapes.

If you are trying to inspect a small mark or a bit of sensor dust right in the far corner of a photo, a standard zoom might actually push that detail off the edge of the screen as the image expands. By having "Zoom Clicked Point to Centre" turned on, Photoshop pulls that corner detail right into your main field of view, making it much easier to work on without having to scroll around.

It really comes down to personal preference. Some people love the control of keeping the image static, while others prefer the software to "hand" them the detail they are looking for by placing it in the middle.

Reality vs Photoshop - Is Faking It Cheating? 🤷‍♂️

Car photography always looks that little bit more dramatic when there's a wet road reflection underneath the vehicle. But what do you do when the road is bone dry? In this guide, I'll walk you through two ways to fake a puddle reflection in Photoshop -- one traditional, one powered by AI -- and then I'll leave you with a question worth thinking about.

Method One: The Manual Approach

Step 1: Select the Car

Start by grabbing the Object Selection tool from the toolbar. In the options bar at the top of the screen, make sure the mode is set to Cloud for the best possible result, then click Select Subject. Photoshop will do a surprisingly good job of selecting the car in just a moment or two.

Step 2: Copy the Car onto Its Own Layer

With your selection active, press Command + J (Mac) or Control + J (Windows) to copy the car up onto a new layer. If you toggle every other layer off, you should see just the isolated car sitting cleanly on a transparent background.

Step 3: Flip It Upside Down

Go to Edit > Transform > Flip Vertical. This flips the car layer to create the basis of your reflection. Now grab the Move tool, hold down Shift (to keep movement perfectly vertical) and drag the flipped car downwards until the tyres of both the original and the reflection are just touching.

If things look slightly off-angle, go to Edit > Free Transform, move your cursor just outside the bounding box until you see the rotation cursor, and give it a gentle nudge until it lines up properly.

Step 4: Add a Black Layer Mask

Rename this layer "Reflection" to keep things tidy. Then, holding down Option (Mac) or Alt (Windows), click the Layer Mask icon at the bottom of the Layers panel. This adds a black mask that hides the layer entirely -- which is exactly what you want for now.

Step 5: Draw the Puddle Shape

Select the Lasso tool and make sure you click directly on the layer mask thumbnail (you should see a white border appear around it, confirming it's active). Now draw a rough, freehand puddle shape beneath the car's tyres -- it doesn't need to be perfect, natural-looking and irregular is actually better here.

Step 6: Fill with White to Reveal the Reflection

Go to Edit > Fill, set the contents to White, and click OK. The reflection will now appear only within the puddle shape you drew.

Step 7: Soften the Edges

Zoom in and you'll notice the puddle edge looks very sharp and unnatural. To fix that, go to Filter > Blur > Gaussian Blur and apply just a small amount -- around 3 pixels is usually enough. This softens the boundary and helps the reflection blend into the ground convincingly.

Finally, you can reduce the opacity of the Reflection layer slightly to make the whole thing look a little more subtle and true to life.

Method Two: Using Adobe Firefly's Generative Fill

If you want a quicker and arguably more realistic result, Photoshop's AI tools can do a remarkable job here.

Step 1: Load the Puddle Selection

Hold Command (Mac) or Control (Windows) and click directly on the layer mask from your first reflection layer. This loads the puddle shape back as an active selection, saving you from having to draw it again.

Step 2: Select the Background Layer

Click on the main image layer, so that Generative Fill works on the background rather than the reflection layer.

Step 3: Run Generative Fill

In the contextual taskbar, click Generative Fill and type a prompt along the lines of: a reflection of car in puddle of water. For the AI model, select Firefly (specifically the Firefly Built and Expand model released in January 2026). If you're on a Creative Cloud Pro account, this won't cost you any credits -- whereas models like Flux or Nano Banana can use anywhere between 20 and 30 credits per generation.

Click Generate.

Step 4: Choose Your Favourite Variation

Firefly will produce three variations for you to compare. Have a look through them and pick the one that looks most convincing. You'll likely notice that the AI does something quite clever: it reflects the sky in the puddle on the far side of the car, just as real water would. Achieving that manually in Photoshop would take considerably more time and effort.

Which Method Should You Use?

For a quick, dirty result, the manual method works well and gives you full control. But for something that genuinely looks like a photograph taken on a wet road, the AI approach is hard to argue with -- particularly because of how naturally it handles the environmental reflections in the water.

A Question Worth Thinking About

Here's something to consider. When photographing that car, there were really two options: bring bottles of water to pour around the car and create a real puddle on the dry road, or add the reflection later in post-production, either manually or with AI.

Both approaches result in a reflection that wasn't originally there. The only difference is when in the process you add it.

So what do you think -- is there a meaningful ethical difference between physically creating something on location and digitally adding it afterwards? When it comes to reflections specifically, does it matter?

Let me know your thoughts in the comments below.

✅ Photoshop JANUARY 2026 - Everything NEW 💥

Adobe dropped Photoshop 27.3.0 on the 27th January, and for once it's not just AI hype and features nobody asked for. This update brings some genuinely useful stuff that photographers and editors have been requesting for years.

Camera Raw tools finally join the party

The headline features are 2 (actually 3) new Adjustment Layers: Clarity & Dehaze and Grain.

If you've ever wanted to use Clarity or Dehaze without opening Camera Raw or converting to a Smart Object, your prayers have been answered. They now work exactly like Curves, Levels or any other adjustment layer. You can mask them, adjust opacity, change blend modes, and they stay fully editable in your PSD.

Clarity is brilliant for adding punch to textures and details in your midtones without blowing out highlights or crushing shadows. Dehaze cuts through atmospheric haze (or adds it if you reverse the slider), and having it as an adjustment layer means you can apply it selectively with a mask.

Grain gets the same treatment. Want to add film-style texture to knock the digital edge off a super-clean file? Chuck a Grain adjustment layer on top, dial it in, and you're done. It's particularly good for black and white work or vintage treatments.

The AI tools are growing up

On the generative side, things have improved quite a bit.​

Generative Fill and Generative Expand now output at up to 2K resolution, which means extended canvases and filled areas look far less mushy and hold detail much better. Adobe has also added model selection, so you can pick the Firefly version that best suits what you're doing.

The real game-changer is Reference Image support in Generative Fill. You can now feed Photoshop a reference photo and it'll try to match the lighting, colour and structure when generating new content. This is massive for compositing work or keeping a series of images consistent.

The Remove tool has also been quietly upgraded. It does a much cleaner job removing objects and people, with fewer obvious smears and repetitive patterns. In most cases you'll get a usable result without needing to follow up with Clone Stamp or Healing Brush.​

Why this one matters

This isn't a flashy update, but it's the kind that actually changes how you work.

Having Clarity, Dehaze and Grain as proper adjustment layers keeps everything inside Photoshop's layer stack where it belongs. No more jumping between Camera Raw, no more Smart Objects eating up file size, no more destructive edits.

The AI improvements make the generative tools feel less like tech demos and more like something you'd actually use in client work. Higher resolution output and better reference matching mean you can rely on them for real projects, not just Instagram experiments.

If you're on Creative Cloud, the update should already be available. The new adjustment layers live in the standard Adjustments panel alongside everything else. Well worth checking out, especially if you shoot landscapes, architecture or do any kind of composite work.​

Photoshop Acting Weird? Fix These 3 Simple Settings ✅

If Photoshop seems to be behaving odd when moving images around and zooming in, in this video I show how to easily fix it and get back control …

🙅🏼‍♂️ How to NEVER forget your Photoshop edits again ✅

I have lost count of the times I have finished an edit, loved the result, and then completely forgotten how I actually got there.

In this video, I am showing you a simple trick using the Photoshop History Log and AI to create a perfect, step-by-step record of every single move you make.

No more guessing which filter you used or what that specific slider value was; it’s like having a digital assistant write your editing recipes for you while you work.

What I cover:

✅ How to turn on the hidden History Log in Photoshop.
✅ Exporting your editing steps as a simple text file.
✅ Using a clever AI prompt to turn that messy log into a clear workflow.
✅ Why this is a game-changer for your consistency and learning.

Come on Adobe 🙏🏻 We NEED THIS FEATURE ⚠️

I've put together this short video because I need to ask a favour from anyone who uses Photoshop Camera Raw or Lightroom. There's a fundamental feature that's been missing for years, and it seriously impacts how we edit our images and the results we achieve.

The Missing Piece in AI Masking

The issue centres on masking, specifically the AI-generated masks available in the masking panel. Being able to select a sky or subject with one click is genuinely incredible, but there's a massive gap in functionality. We have no way to soften, blur, or feather those AI masks after they've been created.

Instead, we're left with incredibly sharp, defined outlines that sometimes look like poorly executed cutouts. This makes blending our adjustments naturally into the rest of the image much harder than it needs to be.

Years HAVE PASSED

Adobe introduced the masking panel back in October 2021. It changed the way we work and represented a huge step forward. Yet here we are, years later, still without a simple slider to soften mask edges.

If you want to blend an adjustment now, you're often stuck trying to subtract with a large soft brush, using the intersect command with a gradient, or employing other crude workarounds to hide the transition. It feels like excessive work for what should be a standard function.

The Competition Gets It Right

What makes this even more frustrating is seeing other software solve this problem elegantly. The new Boris FX Optics 2026 release includes AI masking controls where a single slider softens and blurs the mask outline, and it works incredibly well. Luminar has been offering this functionality for quite a while too.

These tools understand that a mask is only as good as its edges. When the competition provides ways to feather and refine AI selections, the absence of this feature in Adobe's ecosystem feels glaringly obvious.

Adobe's Strengths and Opportunities

Don't get me wrong. I appreciate that Adobe constantly pushes boundaries. We've witnessed tremendous growth over recent years, with developments from third-party AI platforms like Google's Gemini, emerging models, and innovations from Black Forest Labs with Flux and Topaz Labs. It's an exciting time to be a creator.

But I wish Adobe would take a moment to polish what we already have. Adding flashy new features is great, but refining the core workflows we use every single day would be a massive leap forward for all of us.

How You Can Help

Rather than simply complaining about this issue, I've created a feature request post in the Adobe forums. It's been merged with an existing thread on the same topic, which actually helps consolidate our voices into one place.

Here's what I need you to do: click the link below to visit the post and give it an upvote by clicking or tapping the counter number in the upper left. If we can get enough visibility on this, Adobe might finally recognise how much the community wants and needs this feature.

( LINK )

I believe refining existing tools is just as important as inventing new ones. Thank you for taking the time to vote. It really does make a difference when we speak up together.