About Biffer: who we are and what we want

Why we built Biffer

Romain Rissoan has trained and advised companies for many years, particularly on digital tools and artificial intelligence. In this line of work, documents circulate constantly: training agreements, attendance sheets, assessments, meeting reports, contracts, exchanges with HR departments. All of them contain personal data about people who rely on you to keep it confidential, including from an AI model hosted abroad.

The problem arrived with assistants such as ChatGPT, Claude and Gemini. These tools genuinely help to summarize, rewrite, structure and compare. The reflex of "paste the document and ask the question" deserves a step beforehand: the document goes to a third-party server, sometimes retained, sometimes reused for training depending on the settings chosen. For a trainer, a lawyer, an accountant or a doctor, that step beforehand is redaction. ("Biffer" is French for "to strike out".)

The manual method means opening the document, searching for names, replacing them by hand, then going back over headers, tables, comments and metadata. It takes twenty minutes per file, with the risk of missing something that anyone who has done it knows well. We wanted this work to take a few seconds, and to happen on your side.

Biffer is the answer we would have liked to find on the market: desktop software that detects personal data, replaces it with consistent aliases, preserves formatting and lets you restore the original values afterward, all while keeping every byte of the document on your machine. We first built it for our own use, then made it available because the need is shared by every profession that lives on its clients' trust. The how it works page details the result.

Download Biffer

Our principles

Five principles guide every product decision. They are commitments we hold ourselves to, and they explain the shape the software has taken.

• Offline, once the first launch is done. Biffer opens a connection only once, the first time it is launched on a computer, to confirm your email address and receive that computer's key; after that it works entirely offline, processing included. Analysis, replacement and rewriting of the file all happen on your computer. You can use it on a machine cut off from the network, and that is what we recommend for the most sensitive documents.
• Free, with no time limit. An email address, a click on the link you receive, and you can redact 5 files per month with every feature. Your email address is enough: no password, no credit card, and the account serves one purpose only, delivering your keys, free and then unlimited.
• Your usage belongs to you, zero telemetry. Usage statistics, error reports, device identifier: everything stays on your computer. How many documents you process, and which ones, is your business, and that is how we want it. Our privacy policy puts this in writing.
• The code runs on your machine. Which means you can check it yourself: turn off the network, launch Biffer, process a document. Everything works. That test is worth more than any written promise.
• One price, no subscription. You pay €10 excl. VAT (€12 incl. VAT) per computer, paid once, for Mac or PC, and the license stays valid for life for the version you own; only a new major version is paid for again. The software runs on your side, which lets us offer a one-time price rather than a monthly charge. Every feature is available from the free plan onward; our pricing is public.

These principles match what the French data protection authority, the CNIL, recommends in its guidance on AI and the GDPR (in French): minimize the data sent, prefer local processing when possible, keep control over what leaves the organization. Biffer removes a large share of the risk before the question even arises, and leaves you in control of the rest of your compliance.

Download Biffer

What stays in your hands

Honest software also says where its work ends and yours begins. Here is what Biffer leaves to you, so you can decide with full knowledge.

Local OCR, which you check on your own scans. For images and scanned PDFs, Biffer includes character recognition that runs on your computer, then redacts the data it finds by blacking out the pixels, permanently. Recognition quality follows scan quality: on a skewed page, a faint photocopy or handwriting, the highlighted preview lets you spot and add by hand whatever needs it. If your work relies on scanned documents, test the result on your own files: that is what the free plan is for, 5 files per month after you download it.

Legal compliance, informed by Biffer, decided by you. Biffer pseudonymizes: it replaces each piece of data with a stable alias. If you keep the mapping table, the document remains pseudonymized data under the GDPR, because you can return to the original values. If you delete the table, and re-identifying the people becomes unreasonably difficult, you come close to anonymization; an analysis of your context lets you confirm it. Biffer gives you the means to make that decision: it finds what a hurried eye misses, shows you everything in a preview, and you review. You remain responsible for the document that goes out, with much better tools to do so. The personal data and AI page explores this distinction in depth.

Detection you complete in one click. A surname that is also a common word, an unusual first name, an address written in an unexpected way sometimes need your eye, and conversely an ordinary word can look like a name. The highlighted preview is there for this: you add or remove items before saving, and every occurrence follows.

Download Biffer

Our roadmap

We publish our priorities without dates, because we prefer to announce only what we are sure to deliver. The order below reflects what we hear most often from users, and it may change based on your feedback.

1. Windows. The Windows version was the most frequent request: it is available, with exactly the same features as the Mac version (same formats, same detection, same mapping table, same offline operation) and redaction from "Send to" in File Explorer.
2. Better OCR for scans. The built-in character recognition already works offline; we are improving its quality on difficult scans (skewed pages, low contrast, tables) while staying fully offline, true to our principle.
3. Other formats. Presentations, exported emails and OpenDocument files come up often in requests. We will look at them in that order, keeping the rule that applies to current formats: formatting must be preserved.
4. New profession fields. The lists of pre-identified fields for the eighteen professions already covered, from lawyers to social workers, grow as reports come in, and other professions may be added.

Your requests weigh on this order. The report form is for ideas as much as bugs, and every message is read by the person who develops the software. For any other question, the contact page explains how to reach us and how quickly we reply. Biffer is a small project, run by a small company in Lyon, OPCADIA, and that is deliberate: it lets us stay consistent with everything written above.

Download Biffer