Converting confidential documents without the cloud: the 100% local method
Cyrill Semah

In legal, accounting, healthcare or HR, some documents must not leave the machine. A contract, a payslip, a patient file: uploading them to an online service to convert them already means letting them out. Converting confidential documents without the cloud is not a slogan, it's a professional constraint. Here's what actually happens to your files with 100% local conversion, and where the real limits are.
Why cloud conversion is a problem for sensitive documents
Most online converters work the same way: you send the file to a server, it's processed there, you get the result back. For a harmless document, it doesn't matter. For a sensitive one, it creates three concrete problems:
- The transfer itself. As soon as the file is uploaded, it has left your perimeter. What happens to it next depends on a third party.
- Retention. Many services keep files for a while, for caching, error recovery or training.
- Compliance. Professional secrecy, GDPR, internal policies: a transfer to a third party is rarely neutral, and sometimes forbidden.
The only way to rule out these risks is not to send the file at all.
Which professions are most affected
The need to convert without the cloud isn't theoretical, it comes from real constraints:
- Legal. Contracts, filings, case documents. Professional secrecy forbids routing these through an unmanaged third party.
- Accounting and finance. Bank statements, invoices, tax filings. Named financial data that has no reason to leave the machine just to be turned into text.
- Healthcare. Reports, patient files. Health data is among the most sensitive, and its hosting is strictly regulated.
- HR. Payslips, employment contracts, applications. Personal data you don't casually push to an online converter.
The common thread: these professions often want to use AI on their documents without exposing them. Local conversion is what makes the two compatible.
What stays on your machine, what can go over the network
An honestly local conversion must be precise about its boundary. Here is Sygal's, with no ambiguity.
Stays on your machine:
- Your files (Word, Excel, PDF, HTML, email).
- The text converted to Markdown and JSON.
- Your local library and its history.
- Your API keys, encrypted locally.
Can go over the network, only if you trigger it:
- An image or a scan sent to OCR, to the provider you choose, with your own key.
- A URL, if you import one.
In other words, converting text documents is entirely offline. The network only comes into play for two actions you control.
Converting without cloud or account, step by step
The flow is deliberately simple and assumes no sign-up.
- Download the app and open it. No account to create.
- Drop your documents: a file, a whole folder, or a drag and drop.
- Convert locally. For a text document, no network call.
- Export to Markdown or JSON, or keep the result in your local library, searchable anytime.
At no point are your files uploaded. They are read and transformed on your machine.
And OCR on an image? The only point that touches the network
It's better said plainly than hidden: reading an image or a scan needs an OCR engine, and that engine goes through an API. It's the only action that sends anything over the network.
Sygal does it on a bring-your-own-key basis: you choose the provider (OpenAI, Mistral, Claude, NVIDIA NIM, Gemini, or a custom provider) and you supply your own key. Only the relevant image goes out, to the service you decided on, with no markup. If a document is so sensitive that it must let nothing out, two options: skip OCR, or handle it as a text document where possible. The decision stays in your hands, not in an opaque setting.
There's also a simple precaution for images you do send: mask or crop the sensitive area before OCR, so you only transmit what's needed for reading. That way you control not just the provider, but exactly what the image leaving your machine contains.
No account, no telemetry
Privacy doesn't stop at files. There is no account to create, no usage tracking, no silent cloud sync. Sygal never sees your documents, because they never pass through its servers, it doesn't have any for that. For the full detail of what stays local and what can leave, the privacy policy describes the boundary line by line. On compliance, the CNIL, France's data protection authority, remains a solid reference for the GDPR.
Local doesn't mean less practical
Staying on your machine doesn't force you to give up comfort. A local library keeps the history of everything you've converted, with full-text search, without anything being indexed elsewhere. A built-in editor lets you fix the Markdown before export, and batch export handles several files at once. You get the same practical result as an online service, without the hidden price of transferring your data. Local conversion isn't a degraded mode: it's the default one, and the only one that keeps your sensitive documents where they belong.
In short
For confidential documents, the right question isn't "is it encrypted", but "does it leave". A 100% local conversion answers plainly: text documents don't leave the machine, and the only network point, OCR, stays under your control with your key. You can check it on your own files with the 14-day free trial, no card and no account.
Frequently asked questions
That the file is read and transformed on your own machine, without being uploaded to a server. For text documents (Word, Excel, PDF with text), everything happens offline. Nothing is sent, nothing is stored anywhere but on your device.
Try Sygal free for 14 days
Convert PDFs, images and Word docs into clean Markdown, locally on your Mac or PC. No cloud, no subscription.
Download SygalRelated articles

Convert a scanned PDF to Markdown for AI: cost and method
A scanned PDF or an image is expensive for an AI to read. Here's why, and how to convert it to clean Markdown to cut tokens on those specific cases.
Read the article
Preparing your documents for a RAG pipeline: clean Markdown and JSON
A RAG pipeline is only as good as the documents you feed it. Here's how to normalize any format into usable Markdown and JSON, locally.
Read the article
Why we're starting a blog for Sygal
Sygal just went live, without a blog until now. Here's why we're opening one, and what you'll actually find here: real numbers, not generic tips.
Read the article