Here is a question most people never ask when they use an online file converter: where does my file actually go?
With most big converter sites, the answer is: onto a stranger's server. You upload a bank statement, a signed contract, a passport scan. It is processed somewhere, stored for some amount of time (the privacy policies use words like "temporarily" and "up to 24 hours"), and you download the result and hope the deletion promise is true.
We built our converters at MakeHumanAI the other way round, and this post explains why — and why the trade-offs are worth it.
The whole conversion happens in your browser tab
Modern browsers are quietly powerful. Libraries like PDF.js, SheetJS and Mammoth can parse and rebuild documents entirely in JavaScript, on your own machine.
So when someone uses our PDF to Word converter, this is the entire journey of their file:
- They pick the PDF from their device.
- The browser reads it locally and rebuilds it as a document.
- They download the result.
No upload. No server queue. No copy of the file sitting in a data centre in another country. The network tab stays silent because there is nothing to send. You can literally load the page, switch off your internet, and the conversion still works.
The same design runs through the rest of the set:
- PDF to Excel — pulls tables out of statements and invoices into real spreadsheet cells, locally.
- PDF to Image — renders pages to JPG or PNG on the device, which is why it is fast even on a phone.
- The Word and Excel converters going the other way use the same rule: your file, your machine.
Why this matters more than "free" and "no sign-up"
"Free, no sign-up" is on every converter homepage on the internet. It costs nothing to say.
Privacy in this category is different because the files are not random. Nobody converts a PDF of a birthday card. They convert:
- Bank statements, to total them up in Excel
- Signed contracts, to change one clause
- Salary slips, ID documents, visa paperwork
These are exactly the documents you least want copied onto an unknown server as the price of a ten-second conversion. Local processing removes that price entirely — there is nothing to leak, sell, or subpoena, because there is no copy.
The honest trade-offs
Client-side conversion is not magic, and we will not pretend otherwise:
- Very large files are limited by the device. A 200-page scanned PDF will strain a cheap phone in a way it would not strain a server.
- Scanned PDFs stay scans. Turning a photographed page into editable text needs OCR, which is a different (and heavier) job than rebuilding a document that already holds real text.
- The first page load carries the libraries. A few hundred kilobytes of JavaScript, cached after the first visit, in exchange for never uploading your documents. We think that is the right swap.
A ten-second test for any converter
Before you upload a sensitive document anywhere, do this: open the site, open your browser's network tab (or just watch your upload indicator), and convert a test file. If megabytes flow up to a server, your document now exists somewhere else. If they do not — if the work happens quietly in the tab — you keep the only copy.
Documents are personal. The tools that touch them should act like it.
We build free document tools at makehumanai.com — converters, photo and signature resizers, and form-photo checks that run in the browser, without sign-ups and without uploads.
Top comments (1)
那个十秒的网络面板测试是全文最有说服力的设计——隐私承诺看不见，但 network tab 里没有请求就是看得见的证据。补充一个实作细节：解析大 PDF 时记得把 pdf.js 的解析放进 Web Worker，否则主线程卡死会让用户以为页面崩了；再配合分块读入，能显著抬高低端设备的可用上限。trade-off 那段写得很诚实，这类工具能火的真正原因恰恰是这种诚实。