SpeakUp Coach is operated by SpeakUp LLC ("we," "our," or "us"), which provides an AI-powered public speaking coaching platform at speakupcoach.com. We are a website only — we do not publish a mobile or native app. This Privacy Policy explains what data we collect, how we use it, and your rights.

1. Information We Collect

Account Information

When you create an account, we collect:

Email address and password (password is hashed and stored securely by Supabase)
If you sign in with Google — or another sign-in provider we offer — we receive your email address and name from that provider instead, and we never see or store a password for you.
Display name
Language preference (English or Spanish)
Your answers on our welcome form. Right after you create an account, we may ask you to fill in a short welcome form: your date of birth (month and year) and, if you are 18 or older and in a place where we offer contributing, your first language and where you live (your country, and your state if you live in the US). We suggest the country (and, in the US, the state) from your internet connection, but we don't store your IP address for this. Where you live is asked only if you are 18 or older and your internet connection is in a place where we offer contributing (Section 8). In that case the form also asks your gender; you can choose "Prefer not to say". Nobody aged 13 to 17 is asked these questions. New accounts are also asked, optionally, where they heard about SpeakUp; you can skip that question. Accounts created before the welcome form existed can reach the same form from a card on their dashboard.
Why we ask. Your date of birth, because SpeakUp is not for children under 13 and only people 18 or older can contribute recordings (Section 8). Your answer is final; to correct a mistake, contact us. We keep your date of birth only on our own servers: it is never used in analytics or exports and never given to anyone we license recordings to. Your first language, where you live and your gender, for statistics about our users as a group, which never name you; where you live also decides whether we can ask you to contribute (Section 8). Where you heard about SpeakUp, for statistics about how people find us.
Only if you contribute a recording, we copy your first language, your country and state, your age range at the time of the recording and, if you gave them, your gender (the option you picked, never words you typed), the other languages you speak and your accent into that recording's details when you say yes, and the organizations we license it to receive them with it (Section 8). We never do this for anyone who does not contribute.
How long, and changing them. We keep your welcome-form answers while your account exists and delete them when your account is deleted. If we can ask you to contribute (Section 8), in Settings → Your details you can change your first language, where you live and your gender, and in Settings → My contributions you can choose your English accent; otherwise, contact us to change them. You can't change your date of birth. A change applies to recordings you make after it; details already copied into a contributed recording stay as they were when that recording was contributed.

Speech & Practice Data

Audio recordings: When you practice, your speech is recorded and sent to our transcription provider (AssemblyAI) for processing. On our servers, the audio used for coaching exists only for the length of that one request and is never written to our database or to file storage. As soon as AssemblyAI returns the text, we call their deletion endpoint to erase their copy of the transcript — on every take, whether the transcription succeeded or failed. There are two exceptions, described next and in Section 8: if we can ask you to contribute, a full-quality copy that is uploaded to private storage while you decide, and the transcript AssemblyAI returned; and recordings you choose to contribute.
A copy uploaded while you decide: If you are signed in and we can ask you to contribute a recording (Section 8), a full-quality copy of it is uploaded to private storage while you decide, and we use it for nothing else. If you say no ("Not this one" or "Stop asking me"), it is deleted right away. If you don't say yes, it is deleted within 15 minutes. We keep it only if you say yes (Section 8). The transcript AssemblyAI returned for that recording is kept with the copy and deleted at the same time; if you say yes, it is kept with the recording. We do not analyse contributed recordings any further yet. We never make this copy, or keep the transcript, of a recording made without an account, for anyone under 18, after you choose "Stop asking me", or where we don't offer contributing (Section 8). If we ask you whether to contribute and you answer — yes, "Not this one" or "Stop asking me" — we keep a short record of your answer (Section 8).
Transcripts: The full text of your speech — everything you said — is shown to you right after you record. Usually it is not stored: it exists only for as long as it takes to score that recording. What we keep from it is the measurements — overall score, dimension scores, filler word count, words per minute, and duration — plus the short quotations described next. The exception is a recording we can ask you to contribute: its transcript is kept with the copy uploaded while you decide and deleted with it — right away if you choose "Not this one" or "Stop asking me", otherwise within 15 minutes — and if you contribute the recording, we keep its transcript with it (Section 8).
Short phrases you said, quoted inside your feedback: The coaching quotes your own words back to you. The rubric is required to cite exact phrases from your speech as evidence for each judgment, and the top tip either quotes a phrase verbatim or cites one of your measured numbers. Those quotations are saved along with your score. So while we do not keep the full transcript of a recording you don't contribute, we do keep the fragments of it that your feedback quotes.
How your speech was scored: We store the intermediate results behind each score — the measurements we computed, the rubric judgments and the phrases they quote, which AI model produced the score, and whether the AI step succeeded or failed. This exists so a score can be explained and audited rather than taken on trust.
Your first score: One score is kept as a starting point, so later scores can be compared against where you began. It used to come from a diagnostic recording inside a sign-up questionnaire; that flow no longer exists. It now comes from the practice you did on this device before creating an account: when you sign up or sign in, recordings made anonymously in that browser are attached to your account, and the most recent scored one becomes your starting point. If you never practised before signing up, there is no starting point and none is invented.
Time zone: Your time zone is stored so that daily streaks are counted against your local day rather than a server's.

Practice Before You Have an Account

You can record on our public practice pages without an account. To keep practicing after that, we ask you to create a free one. We record the analysis of every recording made there on our server: the measurements, the written feedback, the phrases that feedback quotes, and the practice prompt you spoke on. Those records carry the visitor identifier described below and the page you recorded on — not a name or an email address, because at that point we have not asked you for either.

What we keep from practice without an account. Your audio is sent to our transcription provider for your score and isn't kept on our servers. A recording made without an account is never kept for contributing and can never be contributed.

If you later create an account, that practice follows you into it. This matters, so we would rather spell it out than let the paragraph above imply something it no longer means. When you sign up or sign in, we look for practice recorded in the same browser — matched on the visitor identifier below, and only from the previous 30 days — and attach it to your new account. From that moment those records are associated with your name and email address like any other practice of yours, and the phrases your feedback quotes travel with them. If you never create an account, that association never happens. The match is made on a cookie the browser sends, which identifies a browser rather than a person: if you share a browser profile with someone else, practice they did in it can be attached to your account. Recordings made before you signed in are never kept for contributing and can never be contributed.

We count the free practice takes made in a browser. To enforce the limit on practicing without an account, we keep one record per take that counted, containing: the visitor identifier, the time, how many words you said, the language, and which page it came from. That is the whole record. It holds no audio, no transcript and no phrases from your speech — the word count is a number, not what you said. If a take produced no usable feedback, the record is marked as refunded so it does not count against you. If you go on to create an account, these records are marked as yours, so we can tell how many people the limit actually converted.

Progress & Gamification

We store your lesson progress, XP points, gems earned, streak data, and awards to provide a continuous learning experience. A short form of your display name, with your XP, tier and streak, also appears on the leaderboard, which anyone can see — see Data Storage & Security.

Usage Analytics

We collect usage events (e.g., pages visited, lessons started, recordings completed) to understand how the app is used and improve it. Events are associated with your user ID, a session identifier, and the visitor identifier below. To be accurate rather than reassuring: because these events are linked to your account, they are not anonymous.

Page views. For each page view we also record the address of the page, the page that referred you to it, your browser's user-agent string, any campaign (utm_*) parameters in the link you followed, and an approximate location our host derives from your IP address: a two-letter country code, and the state or province within it. We do not store the IP address itself alongside a page view.

Why we record approximate location, and where we stop. Four reasons. It tells us where the people using SpeakUp actually are, which is how we decide what to build next. And on the practice page it decides which languages we offer first when we ask which language you speak best — a visitor in Tamil Nadu and a visitor in Delhi should not be shown the same list. It is how we suggest your country (and, in the US, your state) on the welcome form (Section 1). And it is one of the two checks we use to decide whether we can offer you contributing recordings (Section 8), which we do not offer in some places: we check both where your connection is and where you told us you live on the welcome form. If you say yes to contributing, two locations are saved with that recording: the country and state you gave on the welcome form, and the country of your internet connection when the recording was contributed, with its state if you are in the US (Section 8). We stop at the state or province. We do not store your city or your coordinates, although our host offers both, because an IP address often points at your internet provider rather than at you, so a more precise figure would look exact without being true. The location from your connection is never shown in the app, except as the country (and, in the US, the state) we suggest on the welcome form, and it is never shared or sold, except as part of a recording you contribute (Section 8). If you ask us for a copy of your data, it is included. If you would rather it were not recorded at all, the contact address at the end of this policy reaches a person. If we have no location for you, we cannot offer contributing.

Visitor cookie. We set a first-party cookie (su_vid) containing a random identifier, so we can tell whether the same browser visited more than once and so we can apply the abuse limits below. It contains no personal information and is not readable by JavaScript on the page. If you later create an account, activity recorded under that identifier before you signed up — such as trying a practice recording on our public demo pages — can be connected to your account. We do this to understand which parts of the site actually help people, not to build a profile of you, and we do not use it for advertising or share it with advertisers.

Abuse limits and IP addresses. Our recording and analysis endpoints work without an account, so we count requests per browser to stop automated scripts from draining them. That counter is normally keyed on the visitor cookie above. When the cookie is absent, it is keyed on your IP address instead, and that address is stored in the counter row. The counters hold nothing else about you.

2. How We Use Your Data

Provide AI-powered speech feedback and coaching
Track your progress and display performance analytics
Power gamification features (XP, gems, streaks, leaderboards, awards)
Generate lesson demonstration audio via text-to-speech
Send you practice emails — a welcome message, and up to three reminders in the week or so after you sign up if you stop practicing
Improve the app based on aggregated usage patterns
Check your age when we ask for your date of birth (Section 1), because SpeakUp is not for children under 13 and only people 18 or older can contribute recordings
Produce statistics about our users as a group from your welcome-form answers, which never name you (Section 1)
Decide whether we can offer you contributing recordings, based on your age and where you live (Section 8)
If we can ask you to contribute, upload a full-quality copy of each recording to private storage while you decide, and delete it unless you say yes (Section 1)
Only if you choose to contribute a recording: keep it and license it to other organizations for AI training, evaluation and research, for which we are paid (Section 8)

About those emails. They are sent through Resend, our email provider. We log whether each one was sent, delivered, opened, clicked, bounced, or unsubscribed from, so we can tell whether they are worth sending and stop mailing addresses that bounce. Every one carries an unsubscribe link, and unsubscribing puts your address on a do-not-send list that we keep so we do not mail you again.

3. Third-Party Services

We use the following third-party services that process your data. As stated in our Terms of Service, by using SpeakUp you agree that your recording and your transcript are processed by the AI providers named here:

Supabase — hosts our database and authentication. All user data is stored in Supabase with row-level security enforced. We also use its file storage for generated lesson audio and, in private storage, for the copy of a recording uploaded while you decide (Section 1) and for recordings you choose to contribute (Section 8). The transcript kept with that copy, or with a contributed recording, is stored in its database. See Supabase Privacy Policy.
AssemblyAI — transcribes your audio recordings into text. Your recording is sent to AssemblyAI, transcribed, and we then call their deletion endpoint to erase their copy of the transcript — both when transcription succeeds and when it fails partway through. For coaching, speaker labelling is off. See AssemblyAI Privacy Policy.
Google — two separate roles. It generates your AI feedback (Gemini): your transcript (not your audio) is sent to Google to produce the written coaching notes on your score. And if you choose to sign in with Google, Google handles that sign-in and tells us your email address and name. See Google Privacy Policy.
Groq — a backup provider for AI feedback. When Google is unavailable or at capacity, your transcript (not your audio) is sent to Groq instead, so you still get a score rather than nothing. Groq no longer transcribes your audio. See Groq Privacy Policy.
Cerebras — a further backup provider for AI feedback, used when both Google and Groq are unavailable. Your transcript (not your audio) is sent to Cerebras, so you still get a score rather than nothing. Cerebras states that it does not retain the inputs and outputs of its inference service. See Cerebras Privacy Policy.
Resend — sends the emails described above. Your email address, your first name and the contents of the message are sent to Resend, which reports back whether the message was delivered, opened, clicked, or bounced. See Resend Privacy Policy.
ElevenLabs — generates text-to-speech audio for lesson demonstrations. Only lesson text is sent; no user data is shared. See ElevenLabs Privacy Policy.
Vercel — hosts the website and provides its own visitor and performance analytics (visits, page views, page load times, Core Web Vitals). The country code we record on a page view comes from Vercel's edge network. See Vercel Privacy Policy.

Recordings you choose to contribute. Contributing is different: it happens only with a separate yes — for one recording, or for that recording and your future ones — never just because you use SpeakUp. We keep the transcript made for your coaching with a contributed recording, and we do not analyse contributed recordings any further yet. Section 8 has the details.

We do not sell or share your personal data with advertisers or data brokers. The one exception is a recording you choose to contribute: we license contributed recordings, and the related information described in Section 8, to AI companies, which pay us for it. Researchers may also use contributed recordings for free. Data brokers or data marketplaces may arrange these licenses, but every licensee signs our contract directly. Some laws call this a "sale" of personal information. Nothing is licensed unless you said yes to contributing it.

4. Data Storage & Security

All data is transmitted over HTTPS/TLS encryption
Passwords are hashed using bcrypt via Supabase Auth
Database access is protected by row-level security: other users cannot read your practice data, scores or progress, and only our own servers can. If you contribute recordings, Section 8 says who else receives them.
The leaderboard is the deliberate exception, and it is public: anyone can load it, including people without an account. It shows a short form of your display name — your first name and the first letter of your last name (for example, "Maria L.") — with your XP, tier and streak, and never your account id. A one-word display name shows only its first run of letters, without digits ("maria2009" shows as "Maria"), and a name with no letters shows as "Speaker". Nothing about your recordings, scores or feedback appears there. If you would rather not be listed by name, pick a display name that does not identify you.
API keys and secrets are stored as server-side environment variables and are never exposed to clients

5. Data Retention

Audio you don't contribute is not kept. Your recording is sent for transcription and discarded once the transcript comes back. The audio used for coaching is not written to our database or to file storage. If we can ask you to contribute, a full-quality copy of your recording is uploaded to private storage while you decide, and deleted right away if you choose "Not this one" or "Stop asking me"; if you don't answer, it is deleted within 15 minutes (Section 1). Once it is deleted, we have no way to play the recording back. Recordings you choose to contribute are kept until you withdraw them or another deletion rule applies — see Section 8 and our Biometric Data Policy.

We do not keep the full transcript of a recording you don't contribute any longer than the recording itself. Usually, once such a recording has been scored, the complete text of what you said is gone. If we can ask you to contribute it, the transcript is kept with the uploaded copy while you decide and deleted with it: right away if you choose "Not this one" or "Stop asking me", otherwise within 15 minutes (Section 1). What survives is the measurements and the short phrases your feedback quotes, as described in Section 1. The transcript of a recording you contribute is kept with it until you withdraw it (Section 8).

Your scores and metrics are kept for as long as your account exists — we do not delete them automatically after a period of time, because they are what your progress chart is made of. If you want a specific practice session removed, contact us and we will delete it.

Deleting your account (see below) permanently removes your welcome-form answers (including your date of birth), your scores, progress, awards and analytics events, including practice that was recorded before you had an account and later attached to it. If you contributed recordings, it also withdraws all of them (Section 8). The few records that outlive an account are listed next.

What deletion does not remove. A few records outlive an account on purpose. None of them contains audio or anything you said.

The count of free practice takes described in Section 1. The record stays; the link to you is erased, so what is left is a word count and a timestamp against a browser identifier with no account attached to it. We keep it because deleting it would hand out a fresh set of free takes, which is the one thing it exists to prevent.
The do-not-send list, if you unsubscribed from our emails (Section 6).
A record of your answers when we asked you to contribute (yes, "Not this one" or "Stop asking me", and turning the Settings switch on or off), and of which details you changed in Settings → Your details (the names of the fields, never what you entered). It is linked to your former account id.
A record that each copy we uploaded while you decided was deleted, which does not name you.
If you contributed recordings: a record of each recording being contributed, and of whether its coaching transcript was added to it, linked to your former account id; a record that each recording was deleted, and the number of each withdrawn take so that it can never be contributed again, neither naming you; and entries in our security log showing who accessed contributed data.

6. Your Rights

Access & export: You can view all your practice data, scores, and progress within the app. If you want a copy of everything we hold about you, contact us and we will provide it. It will not include full transcripts of recordings you did not contribute, because we do not keep them — only the phrases your feedback quotes. If you contributed recordings, the copy includes your contributions, the information we hold about them, and your consent history (Section 8).
Delete a single session: If you would rather a particular recording's score and quoted phrases not be kept, contact us and we will remove that session without affecting the rest of your account. A recording you contributed is withdrawn separately, in Settings → My contributions.
Delete your account: You can delete your account from the Settings page. This permanently deletes your analytics events, practice sessions, awards, progress, your written feedback, and your profile. This action cannot be undone. If you contributed recordings, it also withdraws all of them (Section 8). The records that outlive an account are listed in Section 5 under "What deletion does not remove." One deliberate exception: if you had unsubscribed from our emails, your address stays on the do-not-send list described below, with its link to your account removed. Deleting it would mean we could start emailing that address again, which is the opposite of what unsubscribing asked for.
Withdraw a contribution: In Settings → My contributions you can withdraw one contributed recording, or all of them, at any time (Section 8).
Stop the emails: Every email we send carries an unsubscribe link, and one click stops them. We keep your address on a do-not-send list afterwards, which is the only way to guarantee we do not mail you again.

7. Children's Privacy

As stated in our Terms of Service, you must be at least 13 to use SpeakUp, and SpeakUp is not directed to children under 13. Right after you create an account, we may ask for your date of birth, once, on our welcome form (Section 1). We keep it only on our own servers, to work out your age, and delete it when your account is deleted.

If the answer shows you are under 13, you cannot use SpeakUp. The form shows a short message, "This step isn't available for your account.", and we delete the account promptly — including the email address or sign-in profile details we received, for example through Google sign-in — and the data linked to it, including practice from that browser that had already been attached to it. So that the question cannot simply be answered again with a different date, we keep a marker in that browser (a cookie) for one year; while it lasts, that browser is refused any later date of birth.
If you are 13 to 17, you can use SpeakUp for coaching like anyone else, but you can't contribute recordings while you are under 18: until then we never ask you to, and we never keep a copy of your recordings for contributing. We ask you nothing else on the welcome form except, for new accounts, where you heard about SpeakUp. From the day after your 18th birthday you can contribute like any other adult, but recordings you made before you turned 18 are never contributed.
Accounts created before the welcome form existed have no date of birth on file. They can be asked to fill in the form from a card on their dashboard; until they have, we never ask them to contribute, and they can contribute only if the answer shows they are 18 or older (Section 8). People who practice without an account are not asked their age, and a recording made without an account is never kept for contributing and can never be contributed (Section 8).

We do not knowingly keep or license recordings made by anyone under 18. If we learn that a contributed recording was made before its speaker turned 18, we delete it and, if any organization has received it, tell that organization to delete its copies, as its license requires; we cannot guarantee that it does. If you are a parent or guardian and believe we hold data about a child under 13, contact us at the address below and we will delete it.

8. Contributing Your Recordings (Optional)

This section is about recordings you choose to contribute. If you never contribute, the rest of this policy applies as written. Three things in it can still touch you: the welcome form (Section 1), the copy uploaded while you decide (Section 1), and a short record of your answer if we ask you (below).

What contributing means

Contributing is optional and separate from using SpeakUp. Using SpeakUp never contributes anything, and saying no never changes your coaching, your score or anything else in SpeakUp. It happens only with a separate yes: on the panel we show after your score, for one recording ("Yes, this one"), or for that recording and your future ones ("Yes, this and future ones"); or with the "Contribute future speeches automatically" switch in Settings → My contributions, for your future recordings.
If you say yes, we keep that recording instead of discarding it, and we license it to other organizations — AI companies and researchers — so they can train, test and research AI speech systems. Data brokers or data marketplaces may arrange these licenses, but every licensee signs our contract directly. Our licenses are non-exclusive: the same recording can be licensed to more than one organization. SpeakUp LLC is paid for these licenses. SpeakUp has no paid plan. Licensing contributed recordings is how we plan to cover our costs. You are not paid.
Who can be asked. You must be signed in and 18 or older, and where you live must not be excluded. We check the country and state you gave on the welcome form and the country and state of your internet connection, and if either is excluded, or we cannot tell where your connection is, we do not ask and nothing is uploaded. We do not store your IP address for this. Please don't contribute a speech about your health, religion or other private matters.
What we keep. The recording and its transcript, the details described in Section 1 (your first language, your country and state, your age range at the time of the recording and, if you gave them, your gender, other languages and accent), the country of your internet connection when the recording was contributed (and, in the US, its state), when the recording was made, the type of device, operating system and browser you recorded on, the type of microphone (built-in, headset, Bluetooth or USB; never its name or a device ID) and the recording settings your browser reported, the language you spoke, how long the recording is, and, if you answer the question, where you recorded. You can choose an accent, or leave it unanswered, in Settings → My contributions. We never give the organizations we license it to your email address, your login or your date of birth.
A voice can identify a person. How we treat that is in our Biometric Data Policy.

Withdrawing a contribution

You can withdraw one contributed recording, or all of them, at any time in Settings → My contributions, and deleting your account withdraws all of them. When you withdraw, we stop using and licensing the recording and delete it, with its transcript and the data made from it, from our storage. We keep only a record that it was deleted (Section 5). If a recording has already been licensed, the organizations that received it may keep and use it under their license, whose bans still apply; we cannot promise that they will delete it. AI models that were already trained on a recording cannot unlearn it.

If you choose "Not this one" or "Stop asking me"

The copy uploaded while you decided is deleted at once. We keep a short record of your choice (Section 5). "Not this one" covers only that recording: we may ask you about your next one. "Stop asking me" also means we never ask you again and we stop uploading copies of your later recordings. You can still turn contributing on yourself, with the switch in Settings → My contributions.

Changes

Changes to this policy or to our Contributor Terms that affect recordings you already contributed apply to a recording only if you agree again.

The full documents

The complete text you agree to when you contribute, with every earlier version: SpeakUp Contributor Terms, Privacy Policy: Contributing Your Recordings (Optional) and Biometric Data Policy.

9. Changes to This Policy

We may update this Privacy Policy from time to time. We will notify you of material changes by updating the "Last updated" date at the top of this page. Changes that affect recordings you contributed apply to a recording only if you agree again (Section 8).

10. Contact Us

If you have questions about this Privacy Policy, want a copy of your data, or want to request data deletion, contact SpeakUp LLC at rahul.iyengar@speakupcoach.com.