What Actually Leaves Your Browser When You Compress a File Locally
The file you drop is not posted to a SimplyCompress server. The page around it is not a sealed box. It downloads its own code, it loads ads, and it loads libraries from CDNs, including an English OCR model the first time you ask for one. Those are different facts. Treating them as one slogan is how people either trust too much or dismiss a tool that really does keep the file local.
The file and the page are different traffic
"Nothing is uploaded" is true about the file you drop, and it is easy to hear as "this tab makes no network requests." Those are different claims. The page itself is downloaded from simplycompress.com. Scripts and styles come with it. Ads load from Google. Image, PDF, and code libraries load from public CDNs. Your JPEG, your PDF, and the source you paste are not posted to a SimplyCompress server by this application. You can watch that in DevTools. This article is the list of what you should expect to see, and what you should not.
The architectural comparison with sites that do take the upload, including why a "deleted after one hour" policy is a different product, is already written in Client-Side vs Cloud File Compressors. Read that one for the choice between the two kinds of tools. This one is a field guide to this site.
What the image code does with a photo
You drop a file, or you paste one. The browser hands the page a File object pointing at bytes that are already on your machine. The studio creates an object URL, a blob: URL that only this document can read, and gives that URL to the image element and to Cropper.js. The crop is drawn onto a canvas. Download calls canvas.toBlob, wraps the result in another object URL, and clicks a temporary link with a download attribute. The ZIP path does the same encoding and then lets JSZip build the archive in memory.
There is no fetch of those bytes to simplycompress.com, and no form post. The only fetch in the PDF code is of a data: URL, used to turn a canvas into a blob. It does not leave the tab. I grepped the scripts for fetch, XMLHttpRequest, and sendBeacon before writing this. The image and code scripts do not call them. The PDF script calls fetch only on that local data URL.
You can confirm it on your own machine, which is a better check than trusting a paragraph. Open DevTools, Network panel, enable "preserve log," drop a photo, move the slider, download. Filter by the filename, or look for a POST. You should see the page's own assets and, on a cold load, the libraries below. You should not see your image filename going out as a request. Then turn Wi-Fi off after the page has loaded and encode again. Image Studio still encodes. OCR and the first load of a library will not, because those need the network. That difference is the whole article.
What the page does request
| Request | Why it is there | Sees your file? |
|---|---|---|
| HTML, CSS, and the site's own JavaScript from simplycompress.com | The application. The host can see the usual request logs: IP, user agent, URL, time. The privacy policy says this. | No. The file is not in the request. |
Google AdSense (pagead2.googlesyndication.com) |
Ads, so the tools stay free. Cookies and the page URL, as the privacy policy describes. | Not by anything this page sends. The file is not attached to the ad request. A malicious ad script could read the DOM. That is a different threat from an upload form, and it is why the policy exists. |
| jsDelivr: Cropper.js and exifr, on the image page | Cropping UI, and reading EXIF so the panel can show it. | The libraries run inside the tab. The CDN sees a request for the library, not a request containing your photo, unless the library itself uploaded it. These two do not. |
| esm.sh: pdf-lib, pdf.js and its worker, SortableJS, JSZip, pdf-compress, Tesseract.js, esbuild-wasm | PDF editing, the ZIP, OCR, and the code minifier. esbuild's WASM is loaded from esm.sh/esbuild-wasm@0.20.1/esbuild.wasm. |
Same model. The code asks for the library. A compromised CDN copy could do more, because it runs in the page. That is the trust you extend to any third-party script. |
| Tesseract language data, the first time you check OCR | The English recognition model. The worker is created with eng. |
The download is the model. The scan is passed to the local worker. It is not submitted to a cloud OCR API by this code. If the model cannot be fetched, the suite tells you and continues with a normal export. |
pdf-compress also documents a WASM binary fetched from a CDN by default. Compression still runs locally after that binary arrives. "Local" describes where the PDF is processed. It does not describe a browser with no network.
EXIF, and the checkbox
The metadata panel reads the original with exifr and shows make, model, software, capture date, GPS if present, ISO, and exposure. That read is local. The values are written into the page as text, not as HTML, so a hostile string inside a file does not become markup.
The download does not copy those tags, whether or not the checkbox is ticked. canvas.toBlob builds a new image from pixels. The JPEG APP1 segment, the GPS IFD, and the embedded thumbnail are properties of the old container. They are not in the canvas. I looked at the exporter: the checkbox is stored on the per-image state and restored when you switch thumbnails, and the blob call does not branch on it. There is no "embed the original EXIF" option. Re-encoding is what removes the tags.
So the honest instruction is: the new file does not contain the GPS coordinate you saw in the panel. Verify by opening the download in a viewer that shows metadata, if you are about to send the file somewhere that cares. The original, still in your folder, still contains it. The studio does not delete that original.
A PNG screenshot often has no EXIF. The panel then says no privacy tags were detected and the checkbox disables. That is expected. It does not mean the screenshot is free of private content. The pixels can still show a banking page. Crop those pixels. Metadata scrubbing does not blur a screen.
Where the bytes sit after you are done
- In the tab, until you close it or replace them. Object URLs are revoked as the preview moves on, and the batch holds the
Fileobjects. Closing the tab drops that memory. The site has no account and no file history on a server. - In IndexedDB, briefly, if you click Send to PDF. The handoff stores the already-encoded images in a local database named
SimplyCompressDBso the PDF page can pick them up. That database is on your device. It is not a cloud sync. - In your downloads folder, after you save. That copy is yours. The site cannot delete it. Treat it like any other file you saved.
- In a browser extension's reach, if you installed one that can read page contents. An extension that can see the tab can see the canvas. Local processing does not defend against your own extensions, your own remote-debugging session, or someone looking over your shoulder.
- Not in a SimplyCompress backup. There is nothing to retain. That also means there is nothing we can restore if you close the tab before downloading.
What this does not certify
Local processing is a smaller exposure than uploading a contract to a random converter. It is not a compliance certificate. A clinic, a law office, or a company under an NDA still has to decide whether a page that runs third-party ad and CDN scripts is an acceptable place to open a given file. For some files the answer is yes. For some files the answer is to use a desktop program with no ads and no CDN, on a machine you control. The code cannot make that decision. The privacy policy is the description of ads and logs. This article is the description of the file path. If they ever disagree, the policy and the code are what you should believe, and the code is what you can watch in the Network panel.
PDF metadata that is not EXIF, including the Info dictionary and XMP, is a different container and is covered in Remove Hidden PDF Metadata. Pasting unreleased source into a minifier you do not control would be an upload even though the file is "just text." The minifier on this site runs esbuild in the tab. What it does to the text is in Minifying JavaScript and CSS.
Frequently Asked Questions
Does the EXIF checkbox control whether GPS is removed?
The new file has no EXIF either way. The exporter paints pixels and does not copy the original container. The checkbox is stored with the batch state, and the blob call does not branch on it. The original file on disk still has its metadata.
Does OCR upload the scan?
The code passes the rendered page to a Tesseract worker in the tab. The first run downloads the English model from the library's CDN. That download is the model, not your PDF. If the model cannot be fetched, export continues without OCR.
Can I use the tool offline?
Image encoding works after the page and its scripts have loaded, even if you then disconnect. The first visit, ads, CDN libraries, and the first OCR run need the network. There is no installed offline app yet.
Do you keep a copy I can recover if I close the tab?
No. The site has no account and no file store. Closing the tab drops the in-memory files. Download before you close it. Send to PDF uses a local IndexedDB database on your device for the handoff, not a server.
Compress without handing over the file
Image Studio encodes in the tab. Disconnect the network after the page has loaded if you want to see that the encoder does not need it.
Open Image Studio