Remove Backgrounds Without Uploading Your Photos (Offline Guide)
Every browser-based background remover works the same way: your image leaves your machine, gets processed on someone else's hardware, and comes back — metered by credit and often capped in resolution on the free tier. If that transfer is the problem, you need a tool that cuts out subjects locally. There are four credible ways to do it, and they differ far more than the marketing suggests.
Pick your route in ten seconds
Free, but command line: rembg — Open source, runs locally, no upload. You need Python installed and comfort with terminal flags — and it downloads its model on first run.
Free, but manual: GIMP — Fully offline and free forever. Selection is hand-drawn rather than AI matting, so it is fine for a handful of images and painful at a hundred.
One-time GUI with batch: Formatif AI — Local background removal plus foreground isolation and object removal, chained with upscaling and denoising, across a folder — $12.95 once.
Cloud, when precision matters most: remove.bg — Still the strongest on difficult edges like wispy hair and fur. It uploads your image and meters it by credit — sometimes that trade is the right one.
Full comparison below, or jump to the local workflow.
The short answer
To remove a background without uploading, run the model locally. rembg is the free open-source option — it processes on your machine but requires Python and a terminal. GIMP is free and graphical but manual. Formatif AI is a paid Windows app that adds a GUI and folder batches. All three keep the file local.
That is the whole answer for most people. The rest of this page is about the trade-offs that decide between them — because “local” is not the only variable, and cloud matting is genuinely better on some subjects.
Note on names: Formatif AI (formatif.net) is offline desktop media software, unrelated to Formative, the education platform.
Every option compared
Prices and limits change frequently. Treat the figures as approximate and verify current terms before committing.
| Option | Runs on | Uploads your photo? | Cost | Batch |
|---|---|---|---|---|
| remove.bg (and Canva, Photoroom, Pixelcut) | Their servers | Yes — required | Credits, from roughly $0.20–$0.50/image | API only, metered per call |
| rembg | Your machine | No, after the model is cached | Free, open source | Yes — via command line |
| GIMP | Your machine | No | Free, open source | Manual selection, one image at a time |
| Affinity Photo 2 / Pixelmator Pro | Your machine | No | One-time purchase | Limited — built as editors, not batch tools |
| Adobe Photoshop | Mostly your machine | Only for cloud features such as Generative Fill | Subscription | Actions and scripting |
| Formatif AI | Your machine | No — no cloud tier exists | $12.95 one-time | Yes — folder batches with chained tools |
What actually leaves your machine with a cloud remover
It is worth being precise here, because the privacy argument is usually made badly in both directions. When you use an online background remover, this is what happens:
- Your image is uploaded in full resolution — that is how the service gets the pixels to work with. There is no version of the process that does not transfer the file.
- It is processed on the provider's servers, where it exists as a temporary object on infrastructure you do not control.
- It is returned as a cut-out PNG, usually with the transparency you asked for — and metered against your credit balance.
- The provider deletes it on their own schedule. remove.bg states images are removed automatically after processing; the exact window varies and is their policy, not your control.
All four steps are reasonable engineering. The honest framing is not “cloud tools are unsafe” — it is that the risk is structural rather than malicious. A file that never leaves your disk cannot be retained, subpoenaed, breached or reviewed by anyone, no matter what their terms promise. For client work under NDA, unreleased product shots, medical or legal imagery, and family archives, that distinction is usually the whole reason someone searches for a local option.
The free tier is not really free
It is worth checking what cloud free tiers actually deliver before comparing them to a local tool. remove.bg's free web output is capped at roughly 0.25 megapixels — about 512 × 512 — which is a preview, not a deliverable. Full resolution is credit-based. So the genuine comparison is not “free cloud vs paid local” but “metered cloud vs one-time local”.
The three offline routes, honestly compared
1. rembg — free, local, command line
Works well: Genuinely free and open source. Runs local AI matting with no upload, supports batch folders, and has GPU acceleration available. This is the best free answer if you are comfortable in a terminal.
Limits: Requires Python 3.10+ and a pip install. You drive it with flags rather than a window. It downloads its model on first run, so it is not offline until that has happened. For one image it is fine; for a designer who wants a button, it is friction.
2. GIMP — free, local, manual
Works well: No dependencies, no model downloads, fully offline, and the selection tools are genuinely capable once you know them. Fuzzy Select for high-contrast backgrounds, Foreground Select for harder subjects, then a layer mask and export as PNG.
Limits: There is no AI matting involved — you are selecting pixels by hand. That is acceptable for three images and unworkable for three hundred. It also will not handle hair or fur well without significant effort.
3. Formatif AI — one-time, local, graphical
Works well: A Windows desktop app with Background Removal and Foreground isolation as named tools, plus object removal. Batch a folder, and chain the removal with upscaling and denoising in one run. $12.95 one-time with lifetime updates and no credits.
Limits: Windows 10/11 only, and it needs a DirectX 12 GPU with 6 GB+ VRAM recommended — integration graphics will not run the models. On very fine edges, cloud matting is still ahead.
There is a fourth route worth naming — using an editor you already own. Affinity Photo 2 has a local Select Subject tool with a one-time licence, and Pixelmator Pro does on-device background removal on macOS. Both are excellent editors; neither is designed for unattended folder batches, which is the point at which you start looking for a dedicated tool.
The cost maths
Credit pricing is where the cloud-versus-local decision usually resolves itself, because background removal is a per-image task rather than a monthly one. At roughly $0.20–$0.50 per image:
| If you process | Cloud credits | Local, one-time |
|---|---|---|
| 50 images | ~$10–$25 in credits | $12.95 once |
| 500 images | ~$100–$250 in credits | $12.95 once |
| 5,000 images | ~$1,000–$2,500 in credits | $12.95 once |
The break-even is somewhere around 30 to 60 images. Below that, if you are happy with an upload, credits are the cheaper and simpler answer — and subscription tiers in the $10–$80/month range are entirely sensible for low, steady volume. Above it, or if the upload is unacceptable at any price, local wins on both counts.
And if free is the constraint rather than cheap, rembg does the same job at zero cost, provided you accept the command line. That is the honest comparison for Formatif AI: it is not the cheapest option in this category, it is the one that removes the terminal.
Where cloud still wins
Two situations where uploading is the right call, and pretending otherwise would make this page useless.
Difficult edges: hair, fur, glass, motion
Cloud matting services have invested years in exactly this. Wispy hair against a busy background and translucent objects are where local models still lose. If a dozen portfolio images a month need flawless cut-outs, credits are money well spent.
Zero setup, any machine
A browser and a credit card beat a GPU requirement, a model download and a Python install — especially on a locked-down work laptop, a tablet, or a machine with integration graphics. Local AI is a hardware commitment, and it is fair to weigh that.
Cut out a folder locally, step by step
First, the free route, which is genuinely good if you are technical. Install Python 3.10 or newer, then:
rembg — one image, then a whole folder
pip install rembg rembg i input.jpg output.png rembg p ./input_folder ./output_folder
The model downloads on first run, so connect once. After that it works with the network off.
Second, the graphical route in Formatif AI, where the same operation is a named tool and the batch is a folder:
- Add Background Removal to the workflow and point it at your folder. Subfolders are included, so a directory of product shots does not need flattening first.
- Chain the next step. The usual sequence for listings is background removal, then 2× upscaling to bring the cut-out up to listing resolution, then denoising to clean up compression artefacts the original had.
- Add Foreground isolation if the subject needs separating without a full removal — useful when you want the original background retained as a layer rather than discarded.
- Run the batch. Files process on your GPU with per-file status, so a corrupted image stops itself rather than the queue. Outputs are written back to disk.
The chaining is the part a single-purpose remover cannot do: cutting out 400 product photos is rarely only a cutting-out job. For that specific use case, the bulk product photo workflow walks through it end to end.
Requirement note: Formatif AI needs a DirectX 12 compatible GPU — NVIDIA GTX 900 series or newer, or AMD RX 400 series or newer — with 6 GB+ VRAM recommended. The background removal tool also appears on the Formatif AI toolset.
Removing backgrounds without uploading: FAQ
How do I remove a background without uploading my photo?
Use a tool that runs the model on your own machine. The free option is rembg, an open-source command-line tool: install Python, run pip install rembg, then process a folder with rembg p input output. Paid alternatives with a graphical interface and batch handling include Formatif AI on Windows and Affinity Photo 2. None of them send the image anywhere.
Does remove.bg upload my images?
Yes — it is a cloud service, so your image must be uploaded to their servers for processing. remove.bg states that images are deleted automatically after processing, and that may well be true. The relevant point for privacy-sensitive work is not the retention period but the transfer itself: a file that never leaves your machine cannot be leaked, subpoenaed or retained by anyone, regardless of policy.
Is rembg really free and offline?
Yes, with one caveat worth knowing. rembg is open source and processes locally, so your photos are not uploaded. However, it downloads its AI model on first run, which requires an internet connection once. After the model is cached locally you can disconnect entirely and keep processing. For a fully air-gapped setup, download the model on another machine and copy it across.
What free background removers do not upload your photos?
rembg and GIMP are the two credible free options. rembg gives you AI matting in a terminal; GIMP gives you a graphical interface but manual selection using Fuzzy Select or Foreground Select plus a layer mask. Both are entirely local. Browser-based removers, including the free tiers of remove.bg, Canva and Photoroom, all require an upload by design.
Is the free tier of remove.bg good enough to use?
Only as a preview. The free web tier outputs a low-resolution result capped at roughly 0.25 megapixels — around 512 x 512 — which is too small for print, listings or most social use. Full-resolution output is credit-based, so free is effectively an evaluation rather than a workflow. Check the current terms, as limits change.
Does Photoshop remove backgrounds without uploading?
Usually yes for the Remove Background command, which runs on your machine in current desktop versions. The catch is that Adobe's generative features are cloud-based, so Generative Fill and similar tools do send data to Adobe. If your requirement is that nothing leaves the machine, you have to know which specific feature you are invoking — and that ambiguity is exactly why people choose a tool with no cloud tier at all.
Can local background removal handle hair and fur as well as remove.bg?
Not reliably, and it would be dishonest to claim otherwise. Wispy hair, fur and translucent edges are where cloud matting models have historically been strongest, and remove.bg earned its reputation there. What has changed is that open local models have closed much of the gap on typical subjects — products, people against plain backgrounds, and clean edges — which covers most e-commerce and listing work.
Can I remove backgrounds from a whole folder automatically?
Yes, locally. rembg supports batch folders from the command line with rembg p input output. Formatif AI takes a folder in its interface, applies background removal, and can chain further steps such as upscaling and denoising in the same run, writing outputs back to disk without an upload step.
Keep reading
Cut out a folder without an upload.
Formatif AI handles background removal, foreground isolation, object removal, upscaling and denoising locally, and chains them in one run. It is $12.95 one-time with a 3-day full-featured trial — and the trial runs the whole model set on your own machine.
rembg is free and local and does a real job — if you are comfortable in a terminal, start there and save the money.
Written by
Founder of Formatif, a local-first offline media processing suite for Windows and Linux. Hadi writes about video, image, and audio formats, compression, and privacy-first desktop workflows.