Meta's New AI Creator Tool: Useful, With a Catch
Photoshop's AI Remove Tool Is Finally Offline
Adobe’s Photoshop 27.7 May 2026 release did something photographers have been asking for since the Remove Tool launched: it made the whole thing local. The Remove Tool now runs entirely on your machine, no internet connection required, zero generative credits consumed regardless of your plan. Then in a separate June 15 Creative Cloud update, Lightroom AI culling graduated from early access to general availability — alongside Premiere, After Effects, and other CC app updates. Together, these two releases make for the most meaningful offline photography workflow upgrade Adobe has shipped in years.
There’s a real catch, and we’ll get to it. But the headline is true: Remove, on-device, no cloud call, no credit drain.
Quick Verdict
What changed Details Remove Tool mode On-device (offline capable) Generative credits used Zero, on any plan Model download size ~5GB Mac minimum M1 Max or higher chip, 24GB unified memory Windows minimum NVIDIA RTX 30-series+, 14GB VRAM Lightroom AI culling Out of beta — Face View + Stacking Credit impact None for Remove; Lightroom AI culling is also credit-free Worth updating immediately if: Mac with M1 Max+ and 24GB+ unified memory, or PC with 14GB+ VRAM Blocked for now if: Mac with M1/M2 Pro or standard chip, under 24GB unified memory; or PC GPU below 14GB VRAM
What the On-Device Remove Tool Actually Changes
The Remove Tool is the single most-used AI feature in Photoshop for anyone doing portrait work, product photography, or location shoots. Tourist in your landscape frame, power line over the subject’s shoulder, a brand logo you need cleared before client delivery — Remove handles all of it.
Until 27.7, every Remove Tool operation was a cloud call. You painted over the object, Photoshop pinged Adobe’s servers, the model ran in the cloud, and the result came back. This was fine on a reliable connection. It was a nightmare anywhere else. Event photographers working tethered in venues with flaky hotel WiFi know exactly what the “unable to connect” error looks like mid-culling session. Outdoor photographers in the field can tell you how many times they’ve queued edits for when they get back to a connection.
The credit piece was the other friction. Every generative operation consumed credits from your monthly allotment, and Remove could burn through them fast on complex edits or high-volume shoots. Photographers on lower-tier plans or anyone deep into a project near month-end started rationing Remove usage.
Both problems go away with 27.7’s on-device mode.
You download the model once (Adobe puts it at roughly 5GB) and Remove runs locally from that point forward. No network required. No credits deducted. Photoshop checks your hardware compatibility before offering the download, so you won’t get halfway through a 5GB pull and hit an error.
The result quality in on-device mode sits at parity with cloud Remove for the kinds of edits most photographers run day-to-day. Smaller objects against clean backgrounds, consistent textures, typical portrait cleanup. For complex edge cases like highly detailed objects against cluttered backgrounds or very large removal areas, cloud mode is still available and you can switch between them.
The 14GB VRAM Wall
Here’s where the enthusiasm needs tempering.
The on-device Remove Tool has different hardware floors depending on your platform — and they’re higher than you might expect.
Mac (Apple Silicon): Adobe requires an M1 Max or higher chip with at least 24GB of unified memory. The chips that qualify: M1 Max (32GB+), M2 Max (32GB+), M3 Max, M4 Max, and M4 Pro (24GB+). The M2 Pro doesn’t qualify — it’s a Pro-tier chip, not Max-tier, and an M2 Pro with 16GB falls short on both chip tier and memory. Standard M1, M2, and M3 chips are out.
Windows / NVIDIA: You need an RTX 30-series or higher GPU with a minimum of 14GB VRAM. That qualifies the RTX 3090 (24GB), RTX 4080 (16GB), RTX 4090 (24GB), and RTX 5070 Ti (16GB). Doesn’t qualify: RTX 3080 (10GB), RTX 4070 (12GB), most mid-range cards from the past three or four years. That’s a high-end card requirement.
For event photographers, school portrait shooters, and creators running a mid-tier machine bought two or three years ago, these are real walls. You don’t hit them with typical photo editing work — it’s specifically the AI model’s footprint that demands this much headroom.
This isn’t Adobe being arbitrary. The on-device model is large because it needs to do cloud-quality work without the cloud. You’re trading server resources for a hardware floor on your local machine.
The workaround is simple enough: cloud mode still works exactly as before. If you don’t meet the hardware requirements, nothing breaks — you just stay on cloud Remove with the same credit system as always. The on-device option won’t appear if your hardware doesn’t qualify.
But if you’ve been hoping this update solves the “Photoshop Remove on my 2021 MacBook Pro in the field” problem: it depends entirely on which 2021 MacBook Pro you have.
How the Generative Credits Piece Works (For Everyone, Not Just 14GB Users)
Worth clarifying a detail that’s been muddled in coverage of this update.
The zero-credits benefit applies specifically to on-device Remove operations. If you’re using cloud Remove — either by choice or because your hardware doesn’t meet the on-device requirements — those operations still consume generative credits as before.
So the “no credits” headline is accurate but conditional. On-device mode: zero credits. Cloud mode: credits as normal.
If you’re on Creative Cloud All Apps and running the full 1,000 monthly credit allotment, you’ve probably never hit the limit on Remove alone — it’s the video generation and Firefly image tools that eat credits fast. But if you’re on a plan with a tighter credit ceiling, or if you do high-volume removal work, on-device mode genuinely changes the math.
Photographers who shoot and edit 500+ frames per session should run the math on how many Remove operations they typically do monthly. If you’re on a Photography plan specifically, and Remove has been your most-used credit-consuming tool, the upgrade to a qualifying machine pays for itself differently than it does for someone running occasional edits.
Adobe’s Photoshop 27.7 release notes on the community forum also confirm that on-device mode works regardless of subscription tier. All Plans, not just All Apps. If you have a Photography plan with 14GB VRAM, you get credit-free Remove. That’s a meaningful change for how the feature is positioned.
Lightroom AI Culling Is Out of Beta — And It’s the Other Half of This Story
The June 15 Creative Cloud update was primarily a Lightroom, Premiere, and After Effects story — and the Lightroom side of it matters just as much for photographers. Assisted Culling moved from early access to general availability, and the GA version ships with two capabilities that didn’t exist in the beta.
Face View analyzes eye sharpness on a per-person basis. When you’re shooting groups, standard sharpness analysis evaluates the frame as a whole — a shot where the groom is sharp but the partner is soft might still score as acceptable. Face View breaks this down by individual, so you can flag shots where the subject you actually care about is technically sharp, even if another person in the frame isn’t.
Stacking automatically groups near-duplicate frames. Burst shooters — anyone using continuous drive to catch expressions, action, or light changes — come back from a shoot with dozens of near-identical images. Stacking surfaces the strongest candidate from each burst cluster and groups the rest, so you’re reviewing one decision per cluster instead of comparing twenty nearly-identical shots individually.
Adobe’s claimed reduction in culling time is up to 90%. That number is worth treating skeptically — it’s from Adobe’s own testing, on Adobe’s own datasets, under Adobe’s own methodology. But directionally, the impact on burst shooting is real. PetaPixel’s coverage noted that photographers testing the GA release reported meaningful time savings specifically on event shoots with high frame counts.
For wedding photographers, sports shooters, and anyone doing documentary-style work where you bracket shots and cull heavily, the combination of Face View and Stacking addresses the two most tedious parts of the cull: finding the technically sharp frames and reducing identical-shot overwhelm.
The Assisted Culling features don’t consume generative credits, which puts them in the same bucket as on-device Remove — genuinely useful AI work that doesn’t hit your monthly allotment.
Does This Shift the Offline Photography Workflow?
Taken together, these two updates move Adobe meaningfully in a direction that’s been overdue.
The classic creative workflow pain point for photographers has been that the most time-saving AI features — intelligent removal, smart culling — required cloud connectivity. You could do basic edits offline. But the AI-assisted work that actually compresses editing time was tethered to a server. This forced choices: do client work at home with full connectivity, defer AI editing until you’re back, or accept the cloud dependency as a cost of the workflow.
Lightroom AI culling doesn’t require a network connection for the GA release — the model runs locally. On-device Photoshop Remove doesn’t either (with the VRAM caveat). That’s two significant friction points removed for the subset of photographers who meet the hardware requirements.
The part that’s still cloud-dependent: Photoshop’s generative fill and expand operations, Lightroom’s upcoming Photo to Video feature, most Firefly-powered generation. Adobe hasn’t moved everything on-device and probably can’t — the generation models are significantly larger than inference models for tasks like culling and removal. This is a partial offline upgrade, not a full one.
For the on-device Remove Tool to matter in your workflow, you need to be doing targeted removal — one object, one area, clean it up — not generative extension of canvas or complex scene replacement. Those still go to the cloud. Remove Tool specifically, on-device, for targeted cleanup: that’s the use case that now works offline.
Who This Actually Helps
Event and wedding photographers doing high-volume post-production: the combination of Lightroom AI culling with Face View and the credit-free on-device Remove is a real workflow compression. Cull faster, remove distractions in selects without credit anxiety. If you’ve got a qualifying machine, update immediately.
Location photographers and travel shooters working in areas with unreliable connectivity: on-device Remove means the editing work you used to defer now happens in the field. A 5GB model download once beats a cloud dependency every session.
Studio photographers on Photography plans (not All Apps): on-device Remove means you’re no longer competing with other plan-tier users for credit headroom on the tool you use most. The credits you do have go further.
Anyone running an M1 Max or higher Mac with 24GB+ unified memory (M2 Max, M3 Max, M4 Max, M4 Pro 24GB+), or a PC with an RTX 3090/4080/4090/5070 Ti: the hardware check comes back green. Update, download the model, and stop thinking about the cloud dependency.
Who Has to Wait
M1 and M2 MacBook Pro users: on-device Remove requires M1 Max or higher chip with 24GB unified memory minimum. M2 Pro doesn’t qualify regardless of memory — it’s a Pro-tier chip, not Max-tier. M2 Max with 32GB qualifies; M2 Pro with 16GB does not. Cloud Remove works exactly as before; you’re just not getting the offline/credit-free upgrade yet.
PC photographers running sub-14GB GPU setups: that covers most mid-range cards from the past three or four years. RTX 3080, 3070, most 4070 configurations. The cloud Remove workflow is unchanged; you’re just not getting the on-device upgrade yet.
Anyone expecting this to fix generative fill offline: that’s a different model, a different problem, and Adobe hasn’t addressed it yet. On-device Remove is not on-device fill. These are distinct features. The Firefly AI hub architecture and the broader Creative Cloud AI agent both still operate through cloud infrastructure.
The Bottom Line
The 14GB VRAM floor is a real limitation. A lot of photographers and content creators are going to read this update and discover their machines don’t qualify. That’s frustrating, and it’s not a knock on your gear — it’s just the cost of running a ~5GB inference model locally.
But for the machines that do qualify, this is a genuinely useful shift. Not a headline feature that sounds better than it works. On-device Remove means no connectivity dependency, no credit consumption, and no rationing Remove usage toward the end of the month on tight plans.
Paired with Lightroom AI culling finally hitting GA with Face View and Stacking, this is the kind of update that compounds if your entire photography workflow runs inside the Adobe ecosystem. The cull is faster. The cleanup work doesn’t need the cloud. Two real friction points, addressed.
Go check your hardware. Mac users: you need M1 Max or higher chip with 24GB+ unified memory. Windows users: 14GB VRAM minimum on an RTX 30-series or higher. If you clear those bars, open Creative Cloud, update Photoshop and Lightroom, and download the on-device model. If you don’t — nothing breaks, and the on-device capability will presumably reach more hardware as models get more efficient over time. Adobe has shipped tighter inference models before.
The direction is right. The access is narrower than the headline implies. Both of those things are true.
Adobe shipped Photoshop 27.7’s on-device Remove Tool in the May 2026 release. Lightroom Assisted Culling (with Face View and Stacking) reached general availability in a separate June 15 2026 Creative Cloud update, alongside Premiere, After Effects, and Illustrator. Mac requirements: M1 Max or higher chip with 24GB unified memory minimum. Windows requirements: NVIDIA RTX 30-series or higher with 14GB VRAM minimum. Hardware compatibility is checked automatically before the model download is offered. Verify current requirements at Adobe’s official release notes.
