About
That sentence is the whole company. Everything else on this page is an argument for why it is possible.
The short answer
Photomop is a photo editing suite that runs entirely on your own device. The AI models download to your browser once, cache, and then do their work locally, so your image is never transmitted to us or to anyone else. That single architectural choice is what lets the tools be free to use at full resolution and fast without a queue.
Removing a background is a solved problem. The models that do it are open, well licensed and small enough to fit in a browser tab. Yet almost every tool built on them still asks you to upload the photo, wait in a queue, hand over an email address and then accept a downscaled preview unless you pay.
None of that is a technical requirement. It is a business requirement. A server-side product has to charge for GPU time, so it has to meter you, so it has to identify you, so it has to hold your images. Every friction you have ever hit in a free photo tool traces back to that one decision about where the model runs.
So we moved the model. Your laptop and your phone are both fast enough to run a modern segmentation network. Once the work happens on your side of the wire, the upload disappears, the queue disappears, the cost disappears, and the reason to hold your photos disappears with them. The product gets better in four directions at once, which is usually a sign you have found the right constraint.
The result is a whole studio of tools rather than one clever trick. Cutting a subject out is only the first move; most people then need it on a white plate, at an exact pixel size, under a file size limit, in a format the form will accept. Splitting that across four websites is how the job normally goes. Keeping it in one tab, on one machine, is most of what we are actually building.
A promise not to look at your photos is only as good as the company making it, and companies get acquired, breached and subpoenaed. A photo that was never transmitted is protected by physics instead. There is no bucket of customer images here to leak, because there is no bucket.
Most of the time a cloud editor spends on your photo is not inference. It is the upload, the queue and the download. Delete all three and a warm cut-out comes back in about three seconds, offline, with no spinner in between.
Server GPUs are why cloud photo tools cap a free export at a preview. Our marginal cost per image is zero, so the free plan hands back the full resolution file with no watermark. The business model stops fighting the user.
The economics
Every free tier on the market is a compromise with one line on an invoice. Delete that line and the compromises stop being necessary. Here is the whole ledger, including the side of it that goes against us.
| Line item | A cloud editor | Photomop |
|---|---|---|
| Inference hardware | GPUs rented by the second, one slice per image | The processor already sitting in your device |
| Image storage and transfer | Buckets, lifecycle rules, egress on every download | None, because no image is ever received |
| Identity and metering | Accounts, credit counters and the systems behind them | Nothing to meter, so nothing to build |
| Cost of one more edit | Real, and it grows with every user | Zero, and it stays zero |
| What that forces on the free tier | A resolution cap, a watermark, a credit count or a sign-in wall | Nothing to cap, so full resolution and no watermark |
| What the paid tier has to charge for | The editing itself, because that is the cost | Only the parts with a real bill: API, hosted batch, licensed assets |
Slide the table sideways to see every column.
Read the bottom row and the rest of it follows. A cloud editor has to charge for the editing, because the editing is what costs them. Once each new edit costs us nothing, charging for it stops being a business model and starts being a toll. So the free plan keeps the full resolution and the missing watermark, and Pro is built only out of things that genuinely cost money to run: servers for an API, storage for a batch queue that survives a closed laptop, and photographers or licences for a background library.
It is worth being clear about the shape of this. We are not being generous. We are charging for the parts that have a cost and refusing to invent one for the parts that do not, which is a smaller claim than it sounds and a more durable one.
Status report
Photomop is early. Rather than write the roadmap in the present tense and hope nobody checks, here are the two lists side by side.
No dates attached, because a date we miss is worse than no date at all. The changelog records things on the day they actually ship.
The roadmap
Order is a commitment we can keep. A quarter is a guess in a nicer font. Here is what we are building next and why it is in that position, which is more useful than a timeline that slips.
Each tool gets its own route, opened with the preset the page is about: the exact pixel rules a marketplace publishes, the aspect a platform crops to, the millimetres a passport office measures. Today every card lands on the toolbox, which is the single roughest edge in this build.
Head height, eye line, margin and background colour differ per issuing authority and change without notice. Each country page ships only once the spec has been checked against the authority's own published rules, which is slower than generating sixty of them and the reason we are not.
Batch works in the tab today, which means the tab has to stay open. The version worth paying for survives a closed laptop, and that needs somewhere to live. The hosted API is the same models behind an endpoint, for people who would rather not run them in a browser at all.
Same models, same promise, without the tab. Phones are where most photos start, and a native app gets access to memory and hardware a browser will not hand out. The on-device rule does not bend when these ship: if it ever would, we would not ship them.
House rules
Written down so you can hold us to them, and so a future version of us has to argue with this page before breaking one.
Pro is being built out of things that do not exist yet: batch, the API, the background library. If anything about the free plan changes, the changelog says so in plain words on the day it changes rather than in a diff nobody reads.
No wall in front of the download button, no watermark you have to pay to peel off, no cancellation flow that hides behind four screens. The tools as they stand do not ask who you are, because they do not need to know.
This page, the changelog and the pricing page all say Pro is not on sale yet, because it is not. A roadmap written in the present tense is just a lie with better formatting.
Try it
Open a tool, let the model finish downloading, then turn your wifi off and edit another photo. It still works. That is the entire argument, and it takes about a minute.
Questions this page did not answer go to one inbox, read by the people who wrote it.
Keep reading
The pipeline in five steps, what leaves your device, and how to check it yourself.
Read itThe whole toolbox in three families, with what each one needs before it can run.
Read itWhat shipped, what is still rough, and what is not built yet. Newest first.
Read itWhat the free plan covers today, what Pro adds, and why the split falls where it does.
Read it