Legal

These Terms govern your use of the SonaFact web service at sonafact.com (the "Service"). The provider of the Service ("SonaFact", "we", "us") is identified in the "Service provider" section at the end of this page. By creating an account you accept these Terms. If you do not accept them, do not use the Service.

The Service converts audio recordings into text transcripts and lets you edit, play back, export and translate them, and generate AI documents such as meeting minutes, summaries and task lists.

To use the Service you need an internet connection, a current version of a common web browser with JavaScript and cookies enabled, and an active email address.

• You must be at least 18 years old to create an account.
• You must provide a valid email address and keep your password confidential.
• You are responsible for activity on your account.
• One account is for one person, unless your plan says otherwise.
• Because of sanctions rules, we do not open new accounts for people connecting from Russia, Belarus, Iran, North Korea, Cuba or Syria.

The Service is offered in a free plan and in paid plans with different limits (for example, hours of audio per month). The current plans, prices and limits are shown on the pricing page.

• Paid plans are sold by our reseller, Dodo Payments, which acts as the merchant of record: it processes the payment, issues the invoice and handles taxes. When you buy a paid plan, you also accept the Dodo Payments buyer terms shown at checkout.
• The price is shown before you pay. Applicable taxes, such as VAT, are calculated at checkout based on where you live.
• Your price depends on your account country. We set it when you create your account, from the country you connect from. You cannot change it yourself; if you move to another country, write to support@sonafact.com. Checkout charges the price of your account country and does not let you change the country or the currency.
• A paid plan is paid in advance for a month or a year, as you choose, and renews automatically at the end of each billing period until you cancel it.
• You can change the card a plan is paid from at any time on the Plan page in your account.
• You can pay by card, Apple Pay, Google Pay or a local method shown at checkout for your country, such as UPI, Pix, Kakao Pay, Naver Pay or SEPA Direct Debit. Which methods are offered depends on your country, currency and the amount. In India, renewals paid by UPI or an Indian card are taken about 48 hours after the renewal date, as required by Reserve Bank of India rules. If your account country is India, you can pay by UPI only.
• You can also buy a plan for 30 days or 12 months with a one-time payment. It does not renew; when the time is up, your account moves to the free plan. Buying the same plan again adds the time to the end of your current access. Buying a different plan starts it right away and turns the unused time of the old plan into time of the new one, in proportion to their monthly prices. While a subscription is running, one-time access is not available.

• You can cancel a paid plan at any time on the Plan page in your account. You can also write to support@sonafact.com and we will cancel it for you.
• Cancellation takes effect at the end of the current billing period. Until then you keep all the features of your paid plan; after that your account moves to the free plan.
• To avoid being charged for the next billing period, cancel at least 48 hours before it begins.

• A paid plan starts immediately after payment. Before you pay, you expressly ask us to start providing it straight away and acknowledge that you thereby lose your right to withdraw from the purchase.
• Where mandatory consumer law nevertheless allows you to withdraw, you pay for the part of the billing period you have already used.
• Payments are not refunded, including for the unused part of a billing period, unless the law requires a refund or the Service did not work as described through our fault.
• One exception: if a payment was made at the price of a country other than the one where your card was issued, and the plan costs clearly more in the card's country, we may refund the whole payment and end the plan. Your account country then becomes the country where the card was issued, and you can buy the plan again at that country's price.
• To request a refund, write to support@sonafact.com. Refunds are made by Dodo Payments to the original payment method.

• If you live in Brazil, you can withdraw from the purchase of a paid plan or one-time access within 7 days of the purchase, without giving a reason (Article 49 of the Brazilian Consumer Defense Code).
• If you live in Argentina, you can revoke the purchase of a paid plan or one-time access within 10 days of the purchase, without giving a reason (Article 34 of Law 24.240). The “Botón de arrepentimiento” link at the bottom of our website leads to the withdrawal form, which you can send without an account and without signing in.
• In both cases, the rules in the section above about losing the right of withdrawal and paying for the part already used do not apply to you: we refund the full amount you paid.
• To withdraw, send the form at sonafact.com/withdrawal/ or write to support@sonafact.com from the email address you used for the purchase. If you use the form, we email you a confirmation with the number of your request straight away.
• After you withdraw, we end your paid plan and your account moves to the free plan; your recordings, transcripts and documents stay. We order the refund within 14 days of your request. Dodo Payments returns the money to the payment method you used; how soon it reaches you depends on your bank or card issuer.

• You keep all rights to the recordings you upload and to the transcripts and documents created from them.
• You allow us to store and process your content only to provide the Service to you.
• You are responsible for having the right to record the people in your recordings and to upload the recordings to the Service (see also our Privacy Policy, Section 10).

• upload content that is unlawful or that you have no right to use;
• use the Service to infringe other people's rights, including privacy and intellectual property;
• attempt to break, overload or bypass the security or limits of the Service;
• access the Service by automated means other than those we provide;
• resell the Service without our written consent.

You must not:

We may suspend or block an account that breaches these Terms. Where possible, we will first tell you what the problem is and give you a chance to fix it.

Transcripts, translations and AI-generated documents are produced automatically and may contain errors. Check them before relying on them, especially for legal, medical, financial or other important decisions.

We work to keep the Service available and reliable, but we do not guarantee uninterrupted operation. We may carry out maintenance and change or remove features. We recommend that you export and keep copies of transcripts that are important to you.

To the extent permitted by law, we are not liable for indirect losses, lost profits or loss of data resulting from use of the Service. Nothing in these Terms limits our liability where the law does not allow it to be limited, in particular for damage caused intentionally, and nothing in these Terms limits the statutory rights of consumers.

You can stop using the Service and delete your account at any time in your account settings. Deleting the account ends the contract and deletes your data as described in the Privacy Policy. We may end the contract with at least 30 days’ notice sent by email, or immediately if you seriously breach these Terms. If you are a consumer in the European Union, you have the right to withdraw from the contract within 14 days of creating your account without giving a reason, by deleting your account or sending us a statement by email. This right applies to the free account; for paid plans, Section 5 applies.

You can submit a complaint about the Service to support@sonafact.com. Please describe the problem and, if possible, how you would like it resolved. We will respond within 14 days.

We may change these Terms for important reasons, such as changes to the Service, to the law or to security requirements. We will publish the changed Terms on this page, with the date they take effect, at least 14 days before the changes take effect. If you do not accept the changes, you can delete your account before they take effect.

These Terms are governed by Polish law. If you are a consumer, this does not deprive you of the protection given to you by the mandatory laws of the country where you live, and you may bring a claim before the courts that have jurisdiction under applicable law.

support@sonafact.com

This Privacy Policy explains how SonaFact ("SonaFact", "we", "us") handles personal data when you use the SonaFact web service available at sonafact.com (the "Service"). The provider of the Service and the controller of your account data is identified in the "Service provider" section at the end of this page. For any privacy matter, contact us at privacy@sonafact.com. This policy applies only to the SonaFact web service. It does not apply to the SonaFact desktop application, which stores all data locally on your computer and does not send it to us.

You upload an audio recording. We convert it into a text transcript with timestamps and separate speakers. You can edit the transcript, listen to it, export it, translate it, and ask our AI to produce meeting minutes, summaries or task lists. You can also search all your transcripts at once and ask questions about your recordings.

• email address;
• password, stored only as a one-way cryptographic hash (we never see or store your actual password);
• if you sign in with Google: a Google account identifier (we do not receive your Google password);
• if you sign in with Apple: an Apple account identifier and, if you choose to hide your email address, the relay address Apple gives you instead (we do not receive your Apple password). To withdraw SonaFact's access from your Apple ID when you delete your account, we also keep an encrypted token from Apple and use it for nothing else;
• if you turn on two-step sign-in: the secret from which your authenticator app makes its codes, and one-way hashes of your ten backup codes (each works once) — stored encrypted. We make the QR code for your app on our own server. When you turn two-step sign-in off, we delete them;
• your plan, the date your email was confirmed, the date your account was created, and account status.
• if you buy a paid plan: a customer and subscription identifier from Dodo Payments, the plan you bought and the status of your subscription. We never receive or store your card number. From Dodo Payments we also keep the country of your billing address; while your subscription is running, your price and your daily limit of AI jobs follow it. If you buy one-time access, we also keep each payment's identifier, plan, period, amount, currency and payment method.
• your account country: the two-letter code of the country you connect from when you create your account — or, if we cannot tell it then, the country of your first connection that we can place, or the country where the card for your first payment was issued. We do not store your IP address, only this code. Your prices on the pricing and Plan pages and at checkout, the currency of prices in our emails and your daily limit of AI jobs depend on it. You cannot change it yourself: we change it when you write to us that you have moved, or after a refund described in the Terms (section 5, refunds). We keep it for as long as your account exists;
• if you pay with a card issued in a country other than the one your price was set for: the payment identifier, the country where the card was issued, the country of the price, the plan, the amount and currency of the payment, and how that price compares with the price in the card's country. We use this to check whether the price you paid matches the country of your card (see the Terms, section 5, refunds). We keep it for as long as your account exists;
• if the link through which you come to our website has campaign tags (for example utm_source=newsletter): those tags, saved with your account when you sign up, so that we know which campaigns bring people who use the Service. If you come to our website from another website, we save that website's domain with your account in the same way (for example google.com), never the full address of the page you came from.
• if you cancel a paid plan and tell us why: the reason you chose and your comment.

• audio recordings you upload, together with their file name, size and duration;
• a fingerprint of each file you upload (its size and a hash of its beginning and end), calculated in your browser, so that we can tell you when you upload a file you have already transcribed;
• recordings you make in your browser with the Record button: the sound from your microphone, sent to us in short encrypted parts while you record. When you press Stop — or when no new part has arrived for 10 minutes, for example because the page was closed — we join the parts into one file and delete them, and from then on the recording is handled like one you uploaded. We use your microphone only after you press Record and allow it in your browser, and only until the recording stops. If you record an online call (in Chrome or Edge on a computer), your browser also takes the sound of the tab you choose to share — or, on Windows, of the screen — and mixes it with your microphone into the same recording: only the sound, never the picture. We get that sound only after you choose what to share in your browser, and only until you stop sharing or the recording stops. SonaFact does not join the call; see Section 10 on telling the other people on it;
• recordings you add from a link (for example a file shared from Dropbox or Google Drive, or from your own server): our server downloads the file from the address you give and from then on handles it like a file you uploaded. We keep the file name, not the address. The service the link leads to is one you choose; it is not our processor;
• transcripts, including everything said in the recording, speaker names you enter, and word timings;
• earlier versions of a transcript: before each change you make in the editor, we keep the previous text and speakers, so that you can undo the change or restore an earlier version — stored encrypted like transcripts, at most 50 versions of each transcript, each for 30 days;
• AI-generated documents (minutes, summaries, task lists) and translations;
• highlights you make (the selected text, its colour and your note) and clips (a stretch of time and its name) — stored encrypted like transcripts; the audio of a clip is not stored separately, but cut from your recording when it is played or downloaded;
• videos you upload to have subtitles put into the picture: the video, its name and length, and the look and language of the subtitles you choose. The video is stored in the EU, not encrypted, only for this work: we delete it as soon as the subtitles are in — or after 6 hours if you never start — and we delete the finished video 24 hours after it is ready;
• the names of the folders and tags you create, and which recordings are in them;
• your glossary: names and terms you add, with their variants and language — stored encrypted and sent with your recordings to the processing server in the EU, so that it hears them correctly;
• call scorecards you write (the criteria for analysing a call) — stored encrypted; when you order an analysis, the scorecard is sent with the text of the transcript to the processing server in the EU;
• subtitles you correct in the subtitle editor or upload from your own SRT or VTT file — stored encrypted with the transcript. A video you open in the subtitle editor plays only in your browser and is not sent to us;
• automatic titles: if a recording's file name says nothing about it (for example IMG_4821), after transcription the text of the transcript is sent to the processing server in the EU, which writes a short title from the conversation, and we rename the recording with it. Until you rename the recording yourself, we keep the original file name so that you can restore it. You can turn this off in Settings;
• AI document templates you write (name, purpose, sections, instructions and language) and one-off requests for a document — stored encrypted like transcripts. A document made from a template keeps a copy of it, so it keeps its name after you change or delete the template.

Your content may include personal data of other people who speak in the recording. See Section 10.

To show who said what, the Service distinguishes between the voices within a single recording. It does not identify people by their voice, does not create voice profiles or voiceprints, and does not keep any characteristics of a voice after the transcription is finished. Each recording is processed independently: if the same person appears in several recordings, the Service does not link them. Speakers are labelled generically (for example "Speaker 1") unless you enter their names yourself. If you ask for it, the Service can suggest speakers' names from what was said in the recording — for example when someone introduces themselves — working only from the text of the transcript, never from the voice. It shows the passage where the name is said, and a name is used only if you accept it.

• processing records (job status, processing time, errors) — without the content of your recordings;
• usage counters needed to apply plan limits (for example, minutes used); the number of days on which you used SonaFact and the date of the last one (UTC) — only a number and a date, not what you did — so that we can see whether people come back;
• transcript ratings: whether you rated a transcript good or poor and, for a poor one, the reasons you chose from a fixed list — never any text. Our administrators see only totals. A rating is deleted with its recording. After your first transcription we ask once whether the result is usable (yes, partly or no) and then what was wrong or what you mostly record — answers from fixed lists only, kept like a rating. You can skip the question;
• one-time links sent by email (account confirmation, password reset), stored only as a hash and valid for a limited time;
• technical data such as your IP address, processed by our infrastructure providers to deliver the Service and protect it from abuse;
• your country, for prices: if you are signed in, the pricing and Plan pages show prices for your account country (see section 3, account data). Otherwise our server looks up the country you connect from in the country database kept on the server itself, and only if it cannot tell does it use the two-letter country code that your browser gets from Cloudflare and sends along. We do not store this country. On the other pages of our website, your browser asks Cloudflare for the same code only to show the link for withdrawing from a purchase to visitors from Argentina and Brazil and to suggest your country in the withdrawal form; this code is not sent to our server;
• your country, when you create an account: our server looks up your IP address in a country database kept on the server itself; for this check the address is not sent to anyone. We do this so that we do not open new accounts from countries covered by sanctions (Russia, Belarus, Iran, North Korea, Cuba and Syria). The check does not store your address. With your account we save only two-letter country codes, never the address: your account country (see section 3, account data) and the country of your last connection, refreshed at most once an hour while you use the Service. We use the latter while your account country is not yet known and, when a recording's language is detected automatically, to help choose between Hindi and Urdu script. Both codes are deleted together with your account. IP geolocation by DB-IP;
• website statistics: when you visit sonafact.com, Pirsch counts the visit without cookies — the page address, the website you came from, campaign tags in the link, your browser, operating system, device type, country and screen size. Pirsch uses your IP address only to calculate an anonymous visitor number that changes every day, and does not store it. Pages inside the app and one-time links from emails are not counted. On the pages of our free tools, Pirsch also gets an event when you use a tool or take its result to SonaFact — with the name of the tool only, never your file or text;
• quote checks: a quote you check is compared with your transcripts in our server’s memory and is not saved. Pirsch gets an account event with the result (exact, with differences or not found) and whether you checked one recording or all of them — never the quote;
• account events: a short note that something happened in your account — you signed up, you uploaded a recording (and, if your browser made it smaller before sending, its size before and after), your first transcription or AI document was ready, you reached a limit of the Free plan, you came back after a week, or you bought or cancelled a plan. It goes to Pirsch with the plan, the reason, the type of document or the file format (such as MP3) and your campaign tags, but without your email address, your name, your account number, the file name or anything from your recordings. So that Pirsch can count it together with your visit, it carries the IP address and browser of your last request, which our server keeps in memory for up to 24 hours and does not save. For a purchase, it also carries the amount, the currency and the price level for your country (100, 70, 60, 55 or 50% of the price in US dollars). It also notes that a try without an account was started (with the file format only) and, if it later moved to a new account, that it did. It also notes that you started or saved a recording in your browser (with its length in whole minutes and whether it was interrupted). It also notes that you searched your recordings (with the number of matches only), asked a question about them (with its scope, the number of recordings and whether an answer was found) or clicked a link in an answer — never what you searched for, the question or the answer. It also notes that a link to a recording was created (with the options chosen), that such a link was opened — counted like a visit to our website, without the link or anything from the recording — that a comment was added, and that an account was created from a shared recording. It also notes that a highlight was added (with its colour) or a clip was made (with its length). It also notes that you created an AI document template or ordered a document from a template or from a request in your own words (with your plan only) — never the template, its name or the request. It also notes that a recording was transcribed in the mixed-languages mode (with the codes of its two languages only). It also notes that you went from the Plan page to the payment page (with the plan, the period, the currency shown and your campaign tags). Buttons in our emails carry the campaign tags utm_source=email, utm_medium=lifecycle and the name of the email; if you go from such a button to the payment page, the events of going to the payment page and of the purchase carry the name of that email instead of your account's campaign tags. It also notes that we showed you the “What's next” tips after your first transcription and that you used or closed them (with the name of the step only, such as summary or subtitles) — never the name of the recording. It also notes that you chose several files to upload at once (with how many there were, how many were refused or wait for your limit, and the reasons from a fixed list only) — never their names. It also notes that we showed you a message about a limit of your plan (with the type of limit — minutes, AI, file or other — and the language of the app), that you opened the Plan page (with where you came from: the limit message, the menu, an email, the pricing page or other) and that you pressed a plan's button there (with the plan, the period and the currency shown); for the first two, it also notes whether this happened in the phone app. It also notes your answer to the question after your first transcription (yes, partly or no, and what was wrong or what you mostly record, from fixed lists) or that you skipped it, and whether it was your only transcription. It also notes that only the first 30 minutes of a file longer than the Free plan allows were transcribed, with the file's length as a range from a fixed list (for example 45–60 minutes).
• installing SonaFact on your phone: if you add SonaFact to your home screen from the website, our server reports this to Pirsch once, as an account event, with the system only (Android, iOS or other). If you share a recording to SonaFact from another app on Android, the upload event also notes that it came from the Share menu (with the file format only). Until it is uploaded, the shared file waits in your browser's storage on your device; after 30 minutes SonaFact no longer uploads it and deletes it the next time you open SonaFact. In the SonaFact app for Android, recordings made with the app's Record button and files you share to SonaFact wait in the app's storage on your phone until they are uploaded. The app uses the microphone only after you press Record and allow it on your phone, and only until the recording stops. The app deletes its copy once the upload is finished or cancelled or we refuse the file; a file that has not been uploaded within 7 days is deleted the next time you open SonaFact.

• the file you upload for a try without an account, its name, size and duration, and its transcript — stored and encrypted like all other content;
• a keyed hash of your IP address (not the address itself), so that one internet connection gets one try a day;
• a random code saved in your browser's local storage (not a cookie), which lets this browser show the result and move it to an account;
• the robot check: only when you start a try, our website loads Cloudflare Turnstile, which runs a check in your browser; our server then verifies the one-time result with Cloudflare. For this, Cloudflare processes technical data such as your IP address and browser details, as described in Cloudflare's Turnstile Privacy Addendum.

If you do not create an account within 24 hours, we automatically delete the file, the transcript and the IP address hash; like everything else in our database, the transcript may remain in encrypted backups for up to 30 days (see Section 8). If you create an account in the same browser within 24 hours, the recording and the transcript move to your new account and from then on are kept like your other recordings. A try never runs AI documents and does not count towards the limits of any account.

• the text and timings of the subtitles in a file you translate with our free subtitle translator, and the two languages you choose. Your browser reads the file and sends us only that text and those timings, not the file. We pass them to the processing server in the EU for translation and send the translation back to your browser, which builds the translated file. We do not save the text in our database;
• a keyed hash of your IP address (not the address itself), so that one internet connection gets one file a day, together with a random code that lets your browser fetch the result (stored only as a hash), the languages you chose and the number of subtitles and characters — never the text;
• the robot check (Cloudflare Turnstile), as when you try the Service without an account.

We delete this record after 24 hours. Translating a subtitle file does not create an account and does not count towards the limits of any account.

• the recording you send or forward to our Telegram bot @SonaFactBot — a voice message, an audio file or a video. It first passes through Telegram's servers: Telegram is a service you choose to use, it is not our processor, and its own privacy policy applies to it. Our server then downloads the file from Telegram and processes it exactly like a file uploaded on the website, in the EU;
• your Telegram user ID, the ID of your chat with the bot and the language set in your Telegram app, so that the bot can answer you in your language and keep to the daily limit. The bot does not see your other chats or your contacts;
• if you link the bot to your SonaFact account with a one-time code from Settings: the link between your Telegram user ID and your account. You can unlink it at any time in Settings or by sending /unlink to the bot;
• what the bot sends back — the transcript and, for a linked account, a summary and a link to the editor. These messages stay in your Telegram chat, on Telegram's servers, until you delete them there.

Without a linked account the bot transcribes one file of up to 5 minutes a day for each Telegram account, and we delete that file and its transcript after 24 hours, like a try without an account (they may remain in encrypted backups for up to 30 days, see Section 8). With a linked account, recordings sent to the bot belong to your account, count towards your plan's limits and are kept like any other recording; the summary uses one of your daily AI jobs. Files larger than 20 MB cannot be passed on by Telegram's standard bot service — the bot then asks you to upload the file on the website.

• the email address and the message you enter in the contact form on our website, and the language of the page you write from. Our server passes them on in an email to our support mailbox (support@sonafact.com), sent through Resend, with your email address as the address for our reply. We do not save the message in our database: it is kept in our support mailbox like any email you send us, and we use your email address only to answer you;
• the robot check: our website also loads Cloudflare Turnstile when you send the contact form, and only then. It runs a check in your browser, and our server then verifies the one-time result with Cloudflare. For this, Cloudflare processes technical data such as your IP address and browser details, as described in Cloudflare's Turnstile Privacy Addendum;
• a keyed hash of your IP address (not the address itself), kept only in our server's memory and used only to limit how many messages one internet connection can send in an hour. It is never saved.

We use this data to handle your request (see Section 4, “Handling your requests and complaints”). You do not need the contact form to use the Service: you can always write to support@sonafact.com directly.

• the email address, the country and anything you add in the withdrawal form, and the language of the page you send it from. Our server gives your request a number, passes all of this on in an email to our support mailbox (support@sonafact.com), sent through Resend, and sends a confirmation with the number to the email address you entered. We do not save the request in our database: it is kept in our support mailbox like any email you send us;
• so that we can find your purchase, the email to our support mailbox also contains what our database holds about the account with that email address, if there is one: its plan, the country of its billing address, and the dates and Dodo Payments identifiers of its purchases. The form never tells you whether such an account exists;
• the robot check (Cloudflare Turnstile) and a keyed hash of your IP address kept only in our server's memory to limit how many requests one internet connection can send in an hour — as with the contact form.

We use this data to handle your withdrawal and refund, as consumer law requires (see Section 4, “Handling your requests and complaints” and “Meeting legal obligations”). You can also withdraw by writing to support@sonafact.com directly.

• the questions you ask about your recordings and their answers, with links to the moments in your recordings they are based on — stored and encrypted like transcripts;
• what you search for across your recordings: it is used only to find the matches and is not stored.

Search runs in our server's memory over your transcripts after decryption; we do not keep a separate search index. To answer a question, only the text of the transcripts in the scope you chose is sent to the processing server in the EU — never the audio — and the answer is made only from that text. You can delete each question together with its answer. When you delete a recording, we delete every answer that used it; the question stays in your history with a note that its answer was removed.

• links you create: what the person with the link may see, the expiry date and how many times the link was opened — we store only a fingerprint of each link, never the link itself, and we do not record who opened it;
• comments: the text, the line and moment of the recording it refers to, who wrote it and when it was written or edited — stored encrypted like transcripts;
• the name you show next to your comments, stored encrypted, and whether you want an email about new comments on your recordings (at most one an hour for each recording).

If you share a recording by link, anyone who has the link can open it without an account and see what you chose: the transcript with speaker names and, if you allow it, the audio, selected AI documents, comments and a DOCX or PDF download of the transcript. The audio file itself is not offered for download. If you comment on a recording someone has shared with you, its owner and everyone who has the link can read your comment, signed with the name you chose. Your email address is never shown to anyone.

• keys and connections: if you create an API key in Settings, or connect your AI assistant to SonaFact and allow it access, we keep the name of the key or of the app, the first characters of the key so that you can recognise it, when it was created and when it was last used. We store the key itself only as a one-way hash; an API key is shown to you only once, when you create it;
• what goes out: only what the assistant or program you connected asks for with your key — the list of your recordings (name, date, length and status), the transcripts it asks for, with speaker names and timings, and the places where a word it searches for was said. A connected assistant can only read: it cannot change, delete or share anything. With an API key, your program can also get your email address and plan, download transcripts and subtitles in other formats and add recordings from a link. Recordings in the bin are never shown;
• the AI assistant or program you connect is a service you choose, not our processor: what it does with what it reads from SonaFact is governed by its own terms and privacy policy.

You can disconnect an assistant or delete an API key at any time in Settings → Keys and connections; from then on it no longer works. A key also stops working if your plan no longer includes it, and all keys are deleted with your account.

We do not use advertising, tracking pixels or tracking cookies, and we do not build profiles of visitors. Website statistics are collected without cookies. We set only cookies that are strictly necessary for the Service to work: a session cookie that keeps you signed in and, while you sign in with Google, a short-lived security cookie, which also carries the campaign tags from your link, if there were any. A try without an account keeps its random code in your browser's local storage, not in a cookie. If you install SonaFact on your phone, your browser also keeps a small script (a service worker) that only receives recordings you share to SonaFact from other apps; it does not store pages, transcripts or anything else. When you choose “Sign in to comment” on a shared recording, your browser keeps the address of that recording in its local storage for up to one hour, so that you come back to it after signing in. If you come to our website through a link with campaign tags or from another website, your browser keeps those tags and that website's domain in its local storage for up to 30 days, so that they can be saved with your account if you sign up later. Signing in with Apple uses the same kind of short-lived security cookie, only while you sign in. If you use two-step sign-in, after the first step of signing in a short-lived security cookie waits up to 10 minutes for your code.

We do not use your recordings, transcripts or documents to train AI models. We do not sell your data and we do not share it for advertising.

• the application server and database are located in Poland;
• recordings, exports and backups are stored in the European Union;
• transcription, AI documents and translations are processed on servers in the European Union.
• payments are handled by Dodo Payments, our reseller, which may process your payment data outside the European Union; your recordings, transcripts and documents are never sent to it.

Your account data and everything you upload are processed in the European Union:

We use the following providers to run the Service. Each of them processes data only to the extent needed for its task. Some of these providers are companies established outside the European Economic Area. Where their processing involves a transfer of personal data outside the EEA, we rely on appropriate safeguards required by law, such as the European Commission's Standard Contractual Clauses.

• Recordings are encrypted in storage with AES-256-GCM. Each file has its own key, and that key is itself encrypted with a master key held only on our server.
• A recording is stored unencrypted only while it is being transcribed, which usually takes a few minutes. After transcription it is encrypted. An automatic process checks for any recording that has remained unencrypted for more than six hours and either encrypts or deletes it.
• Transcripts and AI-generated documents are stored encrypted in our database.
• Backups of the database are encrypted before they leave the server and are kept in a separate storage location.
• AI processing: for transcription, the audio is sent to the processing server. For AI documents and translations, only text is sent; the recording itself is not. Processing servers are temporary and do not keep any data after the task is finished.
• Staff access to user accounts is limited to administration and support. Every administrative action on a user account is recorded (who did what and when). These records do not contain the content of transcripts.

No system is perfectly secure, but we design the Service so that your content is protected at every stage.

We keep your recordings, transcripts, AI documents and translations until you delete them or delete your account. If you switch on automatic deletion, we delete only the audio, or the recording with everything made from it, after the period you choose (1 to 365 days from the upload), and we email you 3 days beforehand. Otherwise we do not delete them automatically. When you delete your account, all data linked to it is deleted from our database and storage. Backups: we keep encrypted backups of our database for 30 days. They contain account data, information about recordings, and transcripts and documents. They do not contain audio files. Data you delete may therefore remain in backups for up to 30 days, after which it is permanently removed. Backups are used only to restore the Service after a failure; they are not used to restore individual records, so data you have deleted will not be brought back. We keep the record of your consent to emails with tips and plan information (when and how you gave or withdrew it) for as long as your account exists. A try without an account is the exception: we delete it automatically 24 hours after upload, unless it moves to a new account (see Section 3). We keep your questions and their answers until you delete them or your account; deleting a recording deletes every answer that used it. A shared link works until it expires or you turn it off, and then we delete it. A comment is deleted when its author or the owner of the recording deletes it, together with the recording, and when its author deletes their account. Highlights and clips are deleted with the recording. If you switch on the 7-day bin in Settings, a recording with a ready transcript that you delete goes to the bin first: for 7 days it is hidden from your list, search, questions, shared links, the API and your AI assistant, and you can restore it or delete it for good; after 7 days we delete it for good. With the bin off, Delete removes a recording at once. We delete earlier versions of a transcript after 30 days, and videos with subtitles put into them 24 hours after they are ready.

• access your data and receive a copy of it;
• correct inaccurate data;
• delete your data;
• restrict processing;
• receive your data in a portable format (you can also export your transcripts yourself at any time);
• object to processing based on our legitimate interest.

Depending on where you live, you have the right to:

You can delete recordings, transcripts and your whole account yourself in the Service, or have recordings deleted automatically after a period you choose. For any other request, write to privacy@sonafact.com. We will respond within one month. If you believe we process your data unlawfully, you may lodge a complaint with a data protection supervisory authority — in Poland, the President of the Personal Data Protection Office (Prezes Urzędu Ochrony Danych Osobowych), or the authority in the country where you live or work.

Recordings you upload often contain the voices and statements of other people. You are responsible for having the right to record them and to upload the recording to SonaFact — for example, by informing them or obtaining their consent where the law requires it. When you use the Service for business purposes, we process the content of your recordings on your behalf and according to your instructions. Business customers who need a data processing agreement can request one at privacy@sonafact.com. This also applies to online calls you record: SonaFact does not join the call and does not tell the other people on it that you are recording, so tell them yourself before you start.

The Service is intended for people who are at least 18 years old. We do not knowingly process data of anyone younger.

We may update this policy when the Service changes. We will publish the new version on this page with a new date. If the changes are significant, we will publish them at least 14 days before they take effect.

privacy@sonafact.com

The recording is processed to turn speech into text and to separate what each speaker said.

Within a single recording, the Service distinguishes between different voices so that it can show who said what, for example "Speaker 1" and "Speaker 2". This is done only to structure that one transcript.

Once a transcription is finished, nothing about the voices in it is kept. If the same person speaks in several of your recordings, the Service does not know it is the same person. Every transcription starts from scratch, and one recording has no influence on how the next one is processed.

We do not create voiceprints or voice profiles, and we do not store any characteristics of anyone's voice.

If a transcript shows a person's name, it is because you typed it in.

For AI documents and translations, only the text of the transcript is processed, never the recording.

They are encrypted after transcription and are kept until you delete them or your account, or until the period you set in automatic deletion ends. Details are in our Privacy Policy, Sections 7 and 8.

Before uploading, make sure the people in the recording know about it or have agreed where the law requires this.

privacy@sonafact.com