Privacy Policy

We run no analytics, advertising, or tracking software of any kind. There is no third-party script on this site watching what you do.

Google sign-in is optional; an email address and a password work just as well. If you do use it, we ask Google only for the standard sign-in scopes: your email address, your basic profile (name and profile picture), and your Google account identifier. We request no access to Gmail, Drive, Contacts, Calendar, or any other Google service, and we could not read them if we wanted to.

We use what Google returns for exactly one purpose: creating your Steno account and signing you in. We do not sell it, pass it to anyone, use it for advertising, or train any model on it. If a Steno account already exists under the same verified address, Google sign-in attaches to that account rather than making a second one.

Because Google never shows you our sign-up form, we ask for a display name and a username the first time you arrive. Both are yours to change later in your settings.

OCR runs on hardware we operate ourselves in Hong Kong. Your pages go to a vision-language model we host; they are not sent to OpenAI, Google, Anthropic, or any other model provider.

Translation is on by default. A team's owner or manager can turn it off in that team's settings. While it is on, the recognised text of the pages that team asks to translate, never the page images, is sent to Gemini, a Google model we use through Google Cloud's Vertex AI, which runs outside Hong Kong, and the translations it returns are stored in the team's workspace. Turning translation off stops any further text from being sent; it does not delete translations already stored.

Pages your team reviews to a gold standard can be used to fine-tune a model for that team, and that model serves that team. We do not train a shared or general model on your content, and one team's material never improves another team's results.

When a page you upload is byte-for-byte identical to a page Steno has already read, in your workspace or another team's, and the text recognised on that earlier page is still exactly what the model produced, untouched by anyone, we may copy that machine output into your workspace instead of running the model again. Translation works the same way: when a box's text is exactly the same as text we have already translated, we may copy the stored machine translation instead of sending that text to Gemini. Output of a model fine-tuned for one team is copied only within that team. We never copy anything a person entered or decided, such as corrections, reading order, notes or review status, and the team whose page was copied is not told. A copy is stored in your workspace as your own, stays as it is if the earlier page is later changed or deleted, and counts against your plan exactly like a page processed afresh.

We set one cookie: the session cookie that keeps you signed in. It is httpOnly, so no script can read it, it expires after seven days, and it carries a five-minute cached copy of your session so that every page load need not query the database. There is no advertising or analytics cookie.

Your light or dark theme preference is kept in your browser's local storage and is never sent to us.

The Steno application, its database, and uploaded files run on servers we operate in Google Cloud's asia-east2 (Hong Kong) region. Uploaded files are held either in object storage on those servers or in Google Cloud Storage in the same region; a team's files always live in exactly one of the two, never split between them.

So that the service can recover from a failure, we take hourly backups of the database and of the files in that object storage, stored in Google Cloud Storage in the same region, and hold a standby copy of both, kept continuously up to date on a server we operate ourselves in Hong Kong. Files held in Google Cloud Storage are not copied to that server.

OCR and model training run on hardware we operate ourselves in Hong Kong, which reads a team's pages from that storage over an encrypted connection.

That is the complete list. We do not sell personal data and we do not share it with advertisers or data brokers.

There is no self-service delete button yet. Write to contact@steno-ocr.com from the address on the account and we will do it.

Deleting an account removes its sessions, its sign-in methods, its team memberships, its notes, and its personal workspace with every project, file, and image in it, stored objects included. Work you contributed to someone else's team stays with that team; where a record must survive, your name is detached from it rather than the record being destroyed.

Your name, username, email address, and password are yours to change in your settings at any time.

Ask us for a copy of what we hold about you, ask us to correct it, or ask us to delete it, by writing to contact@steno-ocr.com. Your content itself needs no request: everything you produce in Steno exports in open formats from the app.

We are based in Hong Kong and handle personal data under the Personal Data (Privacy) Ordinance. If you are somewhere that grants you further rights over your data, write to us and we will honour them.

Passwords are stored hashed. Uploaded objects are private and reachable only through an authenticated proxy, never from a public bucket URL. Access to a team's material is re-checked on the server on every request rather than merely hidden in the interface.

No system is faultless. If you find a security problem, write to contact@steno-ocr.com and we will answer.

Steno is a tool for professional and research work and is not directed at children.

If we change this policy we will change the date at the top. If a change materially affects how we handle your data, we will tell account holders by email rather than leaving them to notice.

Questions about this policy, or a request about your data: contact@steno-ocr.com.