Speech Cue Privacy Policy
Last updated 7 October 2026 (Speech Cue Plus subscriptions).
Speech Cue is an iPhone app that helps you learn, rehearse and give a talk. This policy explains what the app does with your information. It is written to be read, not skimmed; it is short.
What the app processes, and why
- Asking first. Nothing is sent to the AI provider until you have agreed in the app, on a screen that lists what is sent. You can see the same list, and withdraw your agreement, in Settings → AI & your data.
- Your slides and notes. When you ask the app to write a script, rewrite a slide, think of likely questions or have Coco, your coach, look at your first slides, and during a presentation or rehearsal, your slides' text and speaker notes (or the words of a speech you typed or scanned), your talk's title, audience and instructions, its plan and your script are sent to our server, which uses an AI model provider (OpenAI). Importing a file sends nothing by itself. Slide pictures are made on your phone and stay there. A printed speech you scan is read on your phone, and the pictures of its pages stay on your phone as your script's pages; they are never sent, only the words, when you ask Coco to look at them.
- Your voice, while you present or rehearse. During a live presentation or a rehearsal, audio from the microphone is streamed to the AI provider (OpenAI) for transcription so the app can follow where you are and give you cues. This audio is processed live and is not recorded or stored by us or, to our knowledge, by the provider beyond processing.
- Your voice, on your phone. When you learn your lines, speech recognition happens on your phone only. Following your place in the on-screen script also uses on-device recognition. In a rehearsal, your voice is also recorded on your phone only so the report can play your lines back to you. You can turn this off before any rehearsal and in Settings, delete the recordings at any time, and they are deleted with the presentation. They are never uploaded.
- Rehearsal transcripts and feedback. What you said in a rehearsal (as text), with delivery numbers worked out on your phone (pace, pauses, filler-word counts), is sent to our server and the AI provider to produce coaching feedback. A transcript is kept on your phone only if you turn that on for a session.
- Coco's voice (once you agree to the AI features). The words of a cue or a line to be spoken are sent to Cartesia, a voice provider, to be turned into Coco's voice (or, if Cartesia is unavailable, to OpenAI). If neither can be reached, Coco stays quiet; nothing is spoken by another voice.
- Your voice in a role play (Practice real situations). While a role play runs, your voice is streamed to OpenAI, or to Cartesia when that engine is chosen, which listens and answers aloud in Coco's voice, together with your talk's text and the line you typed about it when the situation is about them. The transcript is sent once more for the written recap. The recording of the conversation is kept on your phone only. Apart from this and the natural voice, nothing is sent to Cartesia.
- Your calendar (only if you tap Add to calendar). The app can put your practice plan and your talk into your phone's calendar, as events with a reminder. On iPhone it asks for write-only access and reads nothing. On Android it needs calendar permission to find your calendar; it reads only the list of calendars, never your events. Nothing about your calendar leaves the phone.
- Your account (optional). You can use the app without an account. If you sign in with Apple or Google, your name, email address (Apple may give a private relay address instead) and a user ID are kept by Google Firebase Authentication, so the app can show your history of past sessions and reports. Delete the account at any time in Settings → Account → Delete account; your presentations and reports on the phone are kept until you delete them there. If you sign in at the end of the introduction, the practice plan you just made is emailed once to that address through our email provider (Resend), with a PDF of the plan, the first slides' script and Coco's notes attached; they are sent to be delivered and are not kept by us.
- A subscription, if you take one. Apple or Google handles the payment; we never see your card or billing address. The store's signed proof of the purchase (Apple's signed transaction, or Google Play's signed purchase record) is sent to our server, which checks it and keeps, against the same anonymous or account ID, which plan you have and when it ends, so that paid use is not limited. Nothing else about the purchase is kept.
- An anonymous ID, to count usage. When the app first asks our server for something, Google Firebase Authentication gives this installation a random user ID with no name or email attached. It is sent with each request to our server, which keeps, per ID and month, how many requests were made, how many seconds of a rehearsal or presentation were listened to and what they cost us, so that free use can be limited fairly. No content is kept with it. If you sign in, your account's ID is used instead.
- Usage and crash reports (on unless you turn them off in Settings → Privacy). Which features are used (for example that a rehearsal finished and how many minutes it took, how many times Help me was tapped, a score band from learning your lines, whether a talk was a PDF or PowerPoint) and reports of crashes go to Google Firebase with a random app identifier, so we can see what helps and fix what breaks. They never contain your slides, script, words, voice, talk titles or your name. No advertising identifier is used, and nothing is used for advertising.
What we do not do
- An account is optional. Without one we do not collect your name, email or contacts; with one, only your name, email and a user ID, as above. We never see your Apple or Google password.
- We do not sell or share your data with advertisers. The app contains no advertising, does not track you across other apps or websites, and its only analytics are the usage and crash reports above, which you can turn off.
- We do not store your slides, transcripts or audio on our server beyond handling each request. Your presentations, reports and progress are stored on your phone.
Third parties
Our server runs on DigitalOcean (London). AI processing is performed by OpenAI under its API terms, which state that API data is not used to train its models. With the natural voice on, the words to be spoken are voiced by Cartesia under Cartesia's privacy policy. Usage and crash reports are processed by Google Firebase (Google Analytics for Firebase and Crashlytics) under Firebase's privacy and security terms. Apple's on-device speech recognition is used for learning your lines and following your script; no data leaves your phone for that. Without signing in, no name or e-mail is sent: requests to our server carry only the anonymous ID above, and the usage and crash reports carry only a random app identifier and are not linked to an account. Your own name reaches the AI provider only if it is in your slides, notes or script, or, in a role play, as the first name you gave Coco so she can address you.
Your choices
- Choose Not now when the app asks, or turn off Settings → AI & your data → Allow AI features; nothing more is sent, and you can still import slides, write your own script, add cue points and learn your lines.
- Turn off Settings → Privacy → Share usage and crash reports; nothing more is sent to Firebase.
- Deny microphone or speech-recognition permission in iOS Settings; the app then works without the features that need them.
- Delete any presentation, and everything kept with it, from inside the app.
- Delete all data by removing the app.
Contact
Questions about privacy: [email protected].