Effective date: 7 August 2026

Lispr is a voice-to-text app for macOS. This policy explains, in plain terms, what data the Lispr app and the lispr.ai website handle — and, just as importantly, what they don't.

1.Who we are

Lispr is operated by Codebridge Technology, Inc. ("Codebridge", "we", "us"), 8 The Green, Suite 12848, Dover, DE 19901, United States. For any privacy question, contact [email protected].

2.How the Lispr app handles your voice

This is the heart of it, so we'll be specific:

• When you hold the activation key, Lispr records a short audio clip locally on your Mac.
• When you release the key, that clip is sent over an encrypted (HTTPS) connection through our transcription proxy to Groq, Inc., which runs the Whisper speech-recognition model and returns the recognized text.
• The text is inserted at your cursor. That's the whole cycle for plain dictation. If you used the translate key, or you have the formatting pass on, the recognized text makes one more encrypted round trip to a language model to be translated or tidied up, and the result is what lands at your cursor.
• By default we do not store your audio or your transcripts on our servers. Our proxy holds the audio only momentarily, in memory, to forward it for transcription — then discards it. (The exceptions are the two optional sharing settings below, both off unless you turn them on.)
• Dictation never requires an account. You may optionally sign in (section 3) so your plan follows you across devices — but the core of Lispr works without one, and we don't build a profile of you.

Audio sent for transcription is processed by Groq under Groq's own terms and privacy policy. We do not control how Groq processes it — please review Groq's privacy policy.

Optional: sharing your recordings. Settings includes a sharing option — "Share recordings to improve" — that is off by default. If — and only if — you turn it on, Lispr may keep a copy of some of your dictation recordings and their transcripts on our servers and use them to improve transcription accuracy and the speech model. This is the only case in which your audio or transcripts are retained. You can switch it back off at any time in Settings (we stop keeping new recordings immediately), and you can ask us to delete what was already collected by emailing [email protected]. The legal basis is your consent, which you give by enabling the setting and may withdraw at any time.

Optional: fragments containing the word "Lispr". A separate setting — "Share fragments with the word «Lispr»" — is also off unless you accept (the app asks once, after you have used the spoken-assistant feature a few times). If you accept, our server briefly keeps the recording of those dictations while an automated step cuts out the short fragment containing the name and deletes the rest; the fragments are used to teach Lispr to recognise its own name in different voices. You can stop sharing, and delete the fragments already collected, from the app's menu at any time. The legal basis is your consent.

Custom vocabulary and its automatic scan. The Vocabulary feature learns the names, brands and jargon that speech recognition keeps getting wrong. Its automatic scan is on by default: from time to time the app sends recent dictation snippets — each capped at 500 characters — to our server, which extracts candidate words and asks a language model (OpenRouter, section 7) to judge which of them are real names or brands; for each candidate a few short excerpts go with it as context. Those snippets are processed to produce that judgement and are not kept as a transcript archive. You can turn the automatic scan off in Vocabulary… and keep curating the list by hand. Our legal basis is our legitimate interest in making recognition accurate; if you would rather nothing left your machine for this, switch the scan off.

Diagnostics & usage data. By default, the Lispr app sends us technical diagnostics and usage data with each request — such as the app version, the dictation language code, transcription speed (latency), success/error status, microphone type, recording length, and a random per-install identifier. That identifier is not your name and says nothing about you; if you sign in to the optional account, it keeps a list of the installs you signed in on, so from that point the identifier can be tied to your account — that is what lets your plan follow your devices. This helps us fix bugs, keep dictation fast, and decide which languages and platforms to support. It contains no audio and no dictated text. You can turn it off in Settings → Diagnostics (the "Diagnostics & usage stats" switch); when it is off, the app sends none of it. Our legal basis is our legitimate interest in operating, securing and improving the app.

Feedback you send us. The app includes an optional feedback form ("Send Feedback…" in the menu). If you use it, we receive the message you write, the email address you optionally provide for a reply, and — if you leave the "Attach diagnostics" switch on — technical diagnostics: app version, OS version, device model, microphone, system language and the app's technical log, which contains no text of your dictations. Our server also records the country and IP address the report arrived from. We use this solely to handle your report; a copy is delivered to our team as a private notification via Telegram and stored privately until we no longer need it to handle the report. You can ask us to delete a report you sent by emailing [email protected]. The legal basis is our legitimate interest in supporting the app — and you choose what to include in the message.

3.Your Lispr account (optional)

You can create a Lispr account to share your plan and usage allowance across your devices. It is passwordless: you sign in with a one-time code we email you (delivered by Twilio SendGrid), or — when available — with Apple or Google. If you never sign in, this whole section simply doesn't apply to you.

• What the account stores: your email address, sign-in times, your plan and subscription state, per-account counters of dictation minutes (numbers only — computed on our server from the length of the audio it momentarily relays, then the audio is discarded as described in section 2), and the list of devices you signed in on (device name, platform, when it was linked). You can remove a device or sign out at any time from your account page.
• What it never stores: your audio, your transcripts, your dictation history — those stay on your device, exactly as without an account.
• Sign-in sessions: the app keeps its session token in your system's protected storage (macOS Keychain / Windows DPAPI).
• Account identity and the diagnostics switch: when you are signed in, requests from the app carry your account session even if the "Diagnostics & usage stats" switch is off — this is what lets us apply your plan to the request. It adds no diagnostics, no audio and no text; the switch continues to control all diagnostic data exactly as described in section 2.
• Payments: paid plans are sold by Paddle, our merchant of record — Paddle handles checkout, payment details and invoices under Paddle's privacy policy. We receive your plan and subscription status, never your full card details.

The legal basis is the contract with you (providing the account and plan you asked for). You can delete your account and its data at any time by emailing [email protected].

4.macOS permissions

Lispr asks for two macOS permissions, each used only for its stated purpose, all locally on your Mac:

• Microphone — to record your voice while you hold the key.
• Accessibility — to insert the recognized text into the app you're using.

Lispr does not log your keystrokes. To join a new dictation neatly onto text it inserted a moment earlier, it reads a short stretch of text around your cursor (about 256 characters) in the field you're typing in. That reading happens entirely on your Mac, is used only to decide spacing and capitalisation, and is never sent anywhere or written to a log.

5.The lispr.ai website

• Hosting. The site runs on Cloudflare Pages. Cloudflare processes standard request data (such as your IP address and browser type) to serve and protect the site.
• Cookies. The site sets functional cookies — lispr-lang to remember your chosen language, and lispr-consent to remember your cookie choice. Google Analytics (see below) sets analytics cookies; in the EEA, UK and Switzerland these are set only after you accept them in our consent banner. We do not use advertising cookies.
• Language. The site may read your browser's language setting to send you to a page in that language. This is processed in transit and not stored.
• Analytics. We use Google Analytics, a service provided by Google LLC, to understand how the site is used. It collects usage data — such as pages viewed, approximate location, device and browser type, and a truncated IP address — using cookies and similar technologies, so we can improve the site. See Google's Privacy Policy; you can opt out with Google's opt-out browser add-on. Visitors in the EEA, UK and Switzerland are asked to consent before any analytics cookies are set — analytics stays off until you accept. You can change your choice at any time: manage cookie preferences.

6.What we don't do

We don't sell your data. We don't show ads. We don't create advertising profiles. We don't use your audio or transcripts to train any model — unless you turn on the optional "Share recordings to improve" setting (off by default; see section 2).

7.Third parties

• Groq, Inc. — speech transcription, and the language model that translates or tidies up the recognized text when you use those features.
• OpenRouter — the language model that judges vocabulary candidates (only while the Vocabulary auto-scan is on; see section 2).
• Cloudflare, Inc. — website hosting, content delivery, and the infrastructure for the transcription proxy.
• Google LLC — website analytics (Google Analytics).
• Telegram — delivery of feedback and diagnostic notifications to our team.
• Twilio SendGrid — delivery of one-time sign-in codes to your email (only if you use the optional account).
• Paddle — merchant of record for paid plans: checkout, payments, taxes and invoices (only if you buy a paid plan).

8.Retention & international transfers

By default we do not retain your audio or transcripts. The exceptions are the two optional sharing settings (both off by default, see section 2): shared recordings and their transcripts we keep until you ask us to delete them or we no longer need them to improve the service, and shared name fragments until you delete them from the menu. Website request logs are handled by Cloudflare under its own practices. Feedback reports you send us (section 2) are kept until we no longer need them to handle the report, or until you ask us to delete them. Diagnostics and usage data are kept as aggregate technical numbers in our analytics for about 90 days. Error reports — the ones the app sends when a dictation fails, which include its neutral technical log and the IP address the report arrived from — are kept until we no longer need them for debugging. Even with diagnostics switched off, our proxy records the bare technical fact of each request (country, edge location, timing, size, success or failure) so we can run and protect the service, and per-install counters of dictation minutes so we can enforce plan limits. If you created the optional account (section 3), its data — email, plan, minute counters, device list — is kept while the account exists and deleted when you ask us to delete the account. Transcription is performed by Groq, which may process data in the United States or other countries.

9.Your rights

Depending on where you live — for example, the EEA and UK under the GDPR, or California under the CCPA/CPRA — you may have rights to access, correct, or delete personal data, and to object to certain processing. Lispr's dictation works without an account; aside from the two optional sharing settings, we do not keep your audio or transcripts as an archive, and the diagnostics & usage data we collect by default is tied only to a random per-install identifier — so we hold very little personal data about you. You can turn diagnostics off, and if you opted into recordings sharing you can withdraw consent in Settings and request deletion of what was collected. For any request or question, contact [email protected] and we will respond as required by applicable law.

10.Children

Lispr is not directed to children under 13 (or under 16 in the EEA). We do not knowingly collect personal data from children.

11.Changes to this policy

We may update this policy from time to time. When we do, we will revise the effective date above. Material changes will be reflected on this page.

12.Contact

Codebridge Technology, Inc.
8 The Green, Suite 12848, Dover, DE 19901, United States
[email protected]