← Back

Legal

Privacy Policy

Last updated: 16 October 2026

The simple version

When you record yourself, your video (or just your voice) goes to our AI coach (run by Google) so it can write your feedback. As soon as your feedback is ready, we delete the recording. We don't keep it, show it to anyone, or use it to train AI. Google keeps a record of it for up to 55 days, only to check its service isn't being misused, and never uses it to train AI either.

We don't know who you are. There are no accounts and no ads. We count visits with a random number kept in a cookie, only to see how people use the site. It tells us nothing about who you are, and you can switch it off at any time under Analytics choices, at the bottom of every page.

If you choose to give us your email with feedback, we keep it for up to a year, only to reply to you and to tell you about new things worth trying. You can ask us to delete it any time at privacy@stagefright.ai.

Worried about something? Email us, or talk to a parent, teacher or someone you trust.

In plain English: You do not need an account. Your video or audio goes to Google Gemini to create your coaching report. We delete our copy the moment your report is ready, and we ask Google to delete the copy it received straight after. Google keeps a record of what it processed for up to 55 days only to detect misuse of its service, and never uses it to train its AI.

1. Who we are

Stagefright.AI ("we", "us", "our") is an AI-powered public speaking coach operated by Ludo Roupell, trading as Stagefright.AI, a sole trader based in the United Kingdom. Our contact address is 55 Harfield Gardens, London, SE5 8DB.

The data controller for the personal data described in this policy is Ludo Roupell. This policy is governed by UK data protection law: the UK GDPR and the Data Protection Act 2018, as amended by the Data (Use and Access) Act 2025.

You can reach our privacy contact at privacy@stagefright.ai.

2. What we process and why

When you record a rehearsal, your device captures video and audio of you speaking, or audio only if you choose audio mode. Because this can identify you from your face and voice, it is biometric data and is special category data under Article 9 of the UK GDPR, receiving the highest level of legal protection. Your voice alone is still biometric data and is processed and protected in exactly the same way; choosing audio does not make it exempt.

We process your recording transiently and solely to generate your coaching feedback. It is not used for anything else: no advertising, no profiling, no training of AI models, no sale to anyone.

Lawful basis: your explicit consent (UK GDPR Article 9(2)(a) and Article 6(1)(a)), given by ticking the consent box at the record step, before your camera or microphone is switched on. You can withdraw consent at any time by stopping the recording or closing the page. Withdrawal does not affect processing carried out beforehand, and because we keep no copy of your recording, there is nothing of our copy left to delete once your report is ready.

Your report is AI-generated feedback for personal development. It does not produce a legal or similarly significant effect on you under Article 22 of the UK GDPR.

3. Where your recording goes

Your recording takes one short, fixed journey:

  • It is uploaded over an encrypted connection into our EU-based storage (Supabase, Frankfurt, Germany), because the analysis service cannot receive the upload directly. It passes through that storage only while the analysis runs.
  • It is sent securely to Google (via the Gemini API) solely to generate your coaching report. We use Google's paid service, whose terms prohibit Google from using your recording or report to improve or train its products. Google holds the uploaded file only while it is analysed: we ask Google to delete it straight after, and if that request ever fails, Google deletes it within 48 hours. Separately, Google keeps a log of what it processed, including your recording, for up to 55 days, solely to detect and prevent misuse of its service. During that time authorised Google staff may review it if it is flagged. It is then deleted.
  • It is deleted from our storage the moment your report is ready, including when the analysis fails. As a safety net, a scheduled sweep clears anything an interrupted session leaves behind after around 24 hours.

The report itself is kept only in this browser tab and is gone when you close it, unless you print or save it yourself. We keep no copy.

4. What we keep

No accounts, no recordings, no reports. After your session, the only things we hold are:

  • A session-limit cookie: a signed counter (a count, a timestamp, and a signature) that enforces the five-rehearsals-a-session limit. It contains no personal data and expires when your browser session ends. See Cookies and your device.
  • A hashed form of your IP address: kept for around 24 hours for rate limiting and abuse prevention. The address is pseudonymised with a keyed hash before it is stored; we do not store it raw. Lawful basis: our legitimate interests (Article 6(1)(f)) in keeping the service secure, available, and free from abuse.
  • A count of completed rehearsals: each time a rehearsal finishes, we add one line to a tally. The line records four things and nothing else: when it happened, whether it was the first, second, third, fourth, or fifth rehearsal of that visit, whether it was video or audio, and which kind of rehearsal it was. It holds nothing about you and nothing from your recording: no name, no IP address, no cookie value, and no identifier of any kind. For the first 24 hours, the time on a line could in principle be matched to the hashed IP address we keep for rate limiting; after 24 hours that record is deleted and nothing can link a line to anyone. We never make that match. It is not a cookie, no third party sees it, and it does not follow you between visits or between websites. We keep it indefinitely, so that we can see whether the product is being used. Lawful basis: our legitimate interests (Article 6(1)(f)) in understanding whether and how the service is used.
  • A service record of each rehearsal: when it happened, its length, whether it was video or audio, the kind of device (phone, tablet or computer, and the browser and operating-system family), the country it came from, which AI model analysed it, what that cost us, the scores and pace the report gave, the type of talk, how long it took, and whether it worked. It holds nothing from your recording, no name and no IP address. It also carries the random visit-counting number described under Visit statistics, so we can see, for example, how many people who rehearsed came back; we never use it to work out who you are. We keep it indefinitely, to run, price and improve the service. Lawful basis: our legitimate interests (Article 6(1)(f)) in running the service within a sustainable cost.
  • A count of screens reached: as you move through the site, we add one line to a second tally each time a screen is reached. The line records two things and nothing else: when it happened, and which screen it was. For the two questions at the start of a rehearsal (what you most want to get out of it, and whether you rehearse your opening or a random topic), that second thing is the answer you picked. It holds nothing about you: no name, no IP address, no cookie value, and no identifier of any kind. For the first 24 hours, the time on a line could in principle be matched to the hashed IP address we keep for rate limiting; after 24 hours that record is deleted and nothing can link a line to anyone. We never make that match. We can see that a hundred people reached a question and sixty reached the next, which tells us where the site is confusing. We do not follow anyone from one screen to the next or recognise you on a later visit, and two people who reach the same screen, or pick the same answer, produce two identical lines. It is not a cookie, no third party sees it, and it does not follow you between visits or between websites. We keep it indefinitely, so we can see whether changes make the product easier to use. Lawful basis: our legitimate interests (Article 6(1)(f)) in understanding and improving how the service is used.
  • Visit statistics: to see how people use the site and where it is confusing, we keep a short line each time a page is opened, a step is reached (for example, starting a rehearsal or seeing your report) or one of our main buttons is pressed. A line records the time; the page or step; which button, if one was pressed; the website that sent you here (just its name, such as google.com) and any campaign tag in the link you followed; the country your connection comes from (never anything finer); and the kind of device (phone, tablet or computer) with its browser and operating-system family. It also records a random number that we keep in a cookie in your browser (see Cookies and your device), so we can tell how many different people visited and whether someone who tried a rehearsal came back. The number is made up when you first visit and tells us nothing about who you are. For each page you open we also keep how long it was open, how long it was on your screen, and how far down you scrolled. That line holds the page, the kind of device and those three numbers, with no visitor number, so it cannot be linked to you or to your other pages. We store no IP address, no full browser details, nothing you type and nothing from your recording. We use it only to count and to improve the service: never for advertising, never to change what you see, and it is never shared or sold. It is kept in our own database (Supabase, in Frankfurt); no analytics company receives it. Each line is deleted after 13 months. You can switch this off at any time under Analytics choices, linked at the bottom of every page. We then delete the cookie and stop counting this browser. If your browser sends a Global Privacy Control signal, we treat that as switched off from the start. Switching off does not delete lines already kept. Because the number is random and we cannot see which one is yours, we cannot find your lines from your name or email; they are deleted after 13 months. Lawful basis: our legitimate interests (Article 6(1)(f)) in understanding and improving how the service is used. The cookie relies on the statistics exemption in the Privacy and Electronic Communications Regulations: its only purpose is counting, we tell you about it here, and you can object at any time, free of charge.
  • Your email address, but only if you choose to give it: at the end of your report you can leave your email and tick a box, so that we can reply to your feedback and let you know when there is something new worth trying. We only hold an email if you give us one. Leaving it blank changes nothing about your report. We store the address and the exact sentence you agreed to, so we can always show you what you signed up for, and nothing else: no name, no recording, no browsing history. We keep it for up to 12 months from the day you give it to us, then delete it. It is also deleted whenever you ask, either by replying to any email from us or by writing to privacy@stagefright.ai. We do not sell it, share it, or pass it to an advertising service. Lawful basis: your consent (Article 6(1)(a)), given by ticking that box, which you can withdraw at any time.

We do not sell your data, share it with advertisers, or use it for any purpose other than providing and securing the service. We may disclose data where required by law.

5. Cookies and your device

We use no advertising cookies and no cookies from analytics companies. We set at most four cookies. Two are strictly necessary for the service you asked for and hold nothing about you; strictly necessary cookies are exempt from consent under PECR. One counts visits, and you can switch it off; the last only remembers that you did.

  • sf_anon_used: the session-limit counter described above (a count, a timestamp and a signature). It expires when your browser session ends.
  • sf_reviewer: set only if you open a private tester link we sent you. It holds a timestamp and a signature, lifts the rehearsal limit for testers, and lasts one year.
  • sf_vid: a random number used only to count visits (see Visit statistics). It lasts 13 months from your last visit. It is never set if you have switched counting off, or if your browser sends a Global Privacy Control signal.
  • sf_noanalytics: set only if you switch counting off, to remember that choice. It holds the number 1 and lasts 13 months.

The visit-counting cookie is not strictly necessary, so it relies instead on the statistics exemption in the same regulations: we tell you about it here and in the notice on our home page, and you can switch it off at any time with one tap, so we do not show a consent banner.

On your own device: we also store a few small items in your browser so the site works as you expect. They stay on your device and are not sent to us.

Until you close the tab:

  • that you ticked the recording consent box, so we don't ask again before your next rehearsal in the same tab;
  • your setup answers and your report, so the report can be shown to you;
  • if you prepared a real talk, the answers you gave about it, so you can rehearse it again without typing them twice.

Until you clear this site's data in your browser:

  • whether you last chose video or audio;
  • that you closed the cookie notice;
  • if you opened a private tester link, which layout of the report you prefer.

You can clear all of these at any time by clearing this site's data in your browser.

We also use Cloudflare Turnstile, a bot check that protects the service from automated abuse. Its script loads on every page, so Cloudflare receives your IP address, browser details and the page's address when you visit, and when you record we send it your IP address to confirm the check. Cloudflare uses this to protect our service, and also, as a separate controller, to improve its bot detection. See Cloudflare's Turnstile privacy addendum.

Analytics choices

We count visits with a random number in a cookie, only to see how people use the site. Turn this off and we delete that cookie and stop counting this browser.

6. Your rights

You have the right to be informed, to access, to rectify, to erase, to restrict, to data portability, to object to legitimate-interests processing, and to withdraw consent. To exercise any right, email privacy@stagefright.ai. We respond without undue delay and within one month.

In practice there is usually very little to act on. Our copy of your recording is deleted the moment your report is ready, and your report lives only in your browser tab. The one thing we may still hold is your email address, and only if you chose to give it: write to us and it is deleted whenever you ask. We also keep visit statistics, which hold nothing that identifies you; you can switch them off under Analytics choices. You can stop at any time by simply closing the page.

7. How to complain

If you are unhappy with how we have handled your personal information, email privacy@stagefright.ai and tell us what happened. You do not need to use any particular words or form.

We will confirm we have received your complaint within 30 days (usually straight away, by automatic reply), look into it promptly, and tell you what we found and what we are doing about it. If it takes longer than a month we will tell you why and keep you updated.

You also have the right to complain to the UK regulator, the Information Commissioner's Office (ICO), at ico.org.uk/make-a-complaint or on 0303 123 1113. You can do this at any time, but the ICO usually asks people to raise it with us first.

8. Where your data goes

Your recording is stored briefly in the European Union (Frankfurt) while it is analysed. Our storage provider, Supabase, is based outside the UK and may access it from Singapore or the United States for support; that is covered by the UK's approved contract clauses.

Your recording is analysed by Google in the United States, and our website is hosted by Vercel, also in the United States. Our bot check is provided by Cloudflare. All three are certified under the UK Extension to the EU-US Data Privacy Framework, which the UK government recognises as giving adequate protection.

9. Children

Stagefright is designed for adults, but we know students use tools like this to practise presentations, so we have designed it to be safe for younger people too: no accounts, no advertising, nothing that follows you to other websites, and your recording is deleted as soon as your feedback is ready. See "The simple version" at the top of this page.