Initializing secure environment…
Initializing secure environment…
ihatepdf is a set of PDF tools that run in your browser. Your file is opened, processed and written back entirely inside the tab you are looking at, using JavaScript and WebAssembly, so the document itself is never transmitted. That single architectural choice explains the rest: no account, no watermark, no file limit, and it keeps working with the network off. Read the privacy details, the limits, and how the site is funded — including the parts that are inconvenient.
ihatepdf runs every PDF tool locally in your browser using WebAssembly, so your file is never uploaded. There is no account, no watermark and no paid tier, and the site is funded by nothing at all.
It changes what you are allowed to send, not merely what you would prefer to send. A medical record, a contract under NDA, an unredacted draft, a payslip — each becomes a document you can work with rather than a question for a compliance team. A local tool removes the question rather than answering it, and that is a stronger position to be in.
Memory is the binding constraint, and the tiers are detected rather than assumed: a desktop handles around 150 MB per file, a tablet 75 MB, a phone 50 MB. Canvas output is clamped and released as you go. A first OCR or speech run downloads tens to hundreds of megabytes of model weights, shown before the download. None of that is hidden, because a hidden limit is worse than a small one.
Server-side batch processing, a shared team library, a mobile app with push notifications, and someone on call. Those all cost money to run, and without revenue they are not options. What that rules in is a tool that is private by construction, free by necessity, and honest about its limits — which is a smaller product with a real guarantee, rather than a larger one with a privacy policy.
The AI tools send extracted text to a model provider using your own API key. That is a real disclosure and it is the only one in the entire site. It is stated in the tool rather than in a policy page, because the person who needs to know is the person about to press the button. If the document cannot leave your machine in any form, the answer is that these are not the tools for it.
No. PDF processing happens locally in your browser using JavaScript and WebAssembly. Your files never leave your device. That is what makes the site safe for contracts, medical records and financial files — and it is a property of how the site is built, not a promise about its policy.
It is not. There are no ads, no affiliate links, no subscriptions and no paid tier. Nothing is sold and there is no data to sell, because no document and no usage is transmitted. The honest consequence is that there is no budget for large infrastructure, which is exactly why everything runs on your own machine.
Your device's memory, and the speed of its processor. Around 150 MB per file on a desktop, 75 MB on a tablet, 50 MB on a phone, following the device's own detected tier. A first run of OCR, speech or audio transcription downloads a model, with the size shown before it starts. No artificial caps, but real ones.
There is no analytics on your documents and no telemetry about which tools you use. If any measurement is added later, it will load as a script with a src and be covered by a Content-Security-Policy — a deliberate architectural rule rather than a promise, because a promise can be broken quietly and a CSP cannot.
The raw file is never uploaded, in any tool. The AI tools extract text in your browser and send only that extracted text, to the model provider you supply your own API key for, and the request is visible before it is made. If even that is unacceptable for a document, do not use the AI tools on it — use extraction, OCR and reading instead.
Because it is a Progressive Web App with a service worker. After the first page load the application and its assets are cached, and the tools run with no network at all. The AI model weights are precached too, and their download size is disclosed up front rather than discovered mid-operation.
The engines are well-known open-source libraries — pdf-lib, pdf.js, Tesseract, Whisper — all bundled locally rather than loaded from a CDN at runtime. Bundling is a deliberate choice: it removes a third-party supply-chain surface from every page load, at the cost of a slightly larger first load.
Genuinely very large files, since the work is bounded by your memory rather than a datacentre. Multi-user shared libraries, which need a server. Certificate-backed digital signatures, which need a certificate authority. And anything that requires a document to be sent to a server — which is most hosted PDF tools, and is the one thing this site will not do.
More about this site — all free, no upload.