Privacy Policy

Account information: when you sign up, we collect your name, email address, and password credentials (stored securely by our authentication provider). If you personalize your hub, we also store the brand details you provide, such as colors, logo, and voice.

Content you upload: talks, recordings, transcripts, scripts, one-sheets, and any other material you add to the app so that our tools can analyze and improve it for you.

Tool results and history: to keep the tools useful across visits and devices, we save the results they generate for you (for example your SEO/AEO tips, content ideas, and stage lists) to your account, and keep a limited recent history you can return to.

Feature usage counts: some tools include a set number of uses on your plan, such as SEO audits, talk comparisons, audience briefings, briefing podcasts, and voice coaching minutes. For those we store a running count per feature for the current day or month so we can apply the limit that comes with your plan. These are counts only. They do not include what you generated.

Event and audience details: when you use Audience Insider, we store what you tell us about an upcoming engagement: the industry, the organization, the event, the location, who is in the room, your topic, the outcome you want, any notes the organizer sent you, and any links you ask us to research. We also store your speaker profile: your topics, frameworks, books, stories, style, humour, call to action, differentiators, and the subjects you do not speak about.

Optional product analytics: if you choose Allow analytics, we store a random identifier and a closed set of events for landing, signup, onboarding, trial, first value, offers, checkout, and subscription status. If you then sign in, we privately link that random identifier to your account ID so signed payment updates can complete the same funnel. Events contain short allowlisted properties only. They do not contain messages, uploads, generated results, email addresses, full page query strings, raw network addresses, or raw browser descriptions. Raw events are deleted after 13 months.

Contact handoff: FAQ chat is not stored. If you separately submit the human follow up form, we store the email, optional name, optional question, support topic, selected goal or plan when provided, source, and the exact marketing consent choice. These records are separate from chat and access requests.

Conversion safety: when server checkout setup fails, we keep an aggregate event for up to 13 months with the failure stage, requested plan, billing interval, and time. It does not contain an account ID, email, request content, Stripe object, or raw error message.

Payment information: if you subscribe to a paid plan, payments are processed by Stripe. We do not store your full card number. Stripe handles that directly under its own security standards.

To provide the service: transcribe and analyze your talks, generate coaching feedback, power your booking inbox and connected tools, and keep your hub styled to your brand.

To improve the product: understand which tools help speakers most, fix bugs, and prioritize new features.

To communicate with you: send account, security, billing, and product updates. You can opt out of non-essential email at any time.

Optional lifecycle email: existing users are opted out. If you opt in, we store your preference and may queue a small number of messages based on setup, trial, first value, or paid activation state. Every optional message has a signed one click unsubscribe, and withdrawal suppresses queued messages immediately. Stripe remains responsible for billing notices.

Some features send your content to trusted AI providers (such as OpenAI and Anthropic) to transcribe audio and generate analysis, suggestions, and drafts. This processing happens only to deliver the feature you requested.

When you upload a long video to ClipFlow, the original is stored privately with Supabase and sent through temporary, expiring links to our Modal processing workers. Modal uses our speech to text service to produce a transcript and thumbnail, and OpenAI analyzes bounded transcript sections to suggest clips. The original does not pass through our Vercel web server.

When you point ClipFlow at a YouTube video instead, we may download that video once and store it privately with Supabase in the same way, so that exporting clips does not depend on YouTube serving the video again each time. It is kept under the same private access rules and the same 30 day limit as an uploaded original, and it is never made public.

If you request AI video enhancement in ClipFlow, only the selected short clip is processed by the open Real ESRGAN restoration model on our Modal GPU infrastructure. The enhanced file is stored privately with your other rendered clips. We do not use your video or enhanced frames to train that model.

QuickCut, the talk editor, handles a long talk a little differently. When you upload one, the audio is extracted inside your browser and only that audio is sent to our speech to text service, so the recording itself is not sent anywhere a second time. The resulting transcript is then read by OpenAI, which chooses the moments that go into your highlight reel.

When you link a YouTube talk to QuickCut instead, we read only that video’s public details and captions from YouTube. We do not download or store the video. Our Modal workers fetch it only at the moment you export a reel from it, and the reel they produce is what gets stored, not the source.

Choosing AI 1440p or AI 4K in QuickCut sends that reel through the same open Real ESRGAN restoration model on our Modal GPU infrastructure. The finished file is stored privately in your account. We do not use your video or enhanced frames to train that model.

The Speech Analyzer works a little differently again. When you analyze a talk, your recording is transcribed by an open speech-to-text model (NVIDIA's Parakeet, or OpenAI's Whisper as a fallback) and then run through TRIBE v2, an open research model from Meta that predicts how a listener's brain is likely to respond. These models run on third-party GPU infrastructure (Modal), which processes your recording only to produce your report. The results are model estimates meant to help you rehearse. They are not measured brain activity, and not medical information.

We do not sell your content, and we do not use it to train our own advertising profiles. AI providers process the content under their own terms and applicable data-processing agreements.

Our home page ends with a message box where anyone, signed in or not, can ask a question about Speaker Hub. This is the only feature on the site you can use without an account, so it is worth saying exactly what happens.

What you type is sent to our AI provider (OpenAI) along with a description of the site, so it can write an answer. That description is public marketing information: our tools, our plans, our prices, and the questions and answers already published on the page. Nothing about any speaker or any account is included.

We do not save the conversation. It is not written to a database, it is not attached to an account, and it disappears when you close or reload the page. We do keep a short lived, hashed record of the network address a message came from, purely so we can limit how many questions one visitor can ask in a few minutes and stop the feature being abused. That record does not contain your message.

The assistant only answers questions about this website. It has no access to any account, any talk, any booking, or any other data in the product, and it cannot make changes on your behalf. Please do not type anything confidential into it.

Audience Insider researches the industry you are about to speak to and then lets you rehearse out loud. It handles more kinds of data than our other tools, so here is exactly what happens.

Research: the event and audience details you enter, plus any links you add, are sent to our AI provider so it can search the web and gather what is publicly documented about that industry. Results are stored on your account with the sources they came from. If web research is unavailable, we still write the briefing from what you told us, and we label it clearly so you know nothing in it is sourced.

Optional research providers: if the deployment is configured for them, we may also send your industry, location, and research questions to Tavily (web search) or Google Data Commons (public statistics). These are off unless keys are set, and the tool works fully without them.

Podcast audio: when you ask for the narrated version, we send the briefing text to our AI provider to generate speech. The resulting audio file is stored privately on your account and is only ever served to you through a temporary, expiring link. It is not public and is not shared.

Microphone audio: we do not store recordings of your voice. In the live voice mode, your microphone streams directly to the AI provider for the length of the session and is not saved by us. In the recorded-message mode, your clip is sent for transcription and discarded once the text comes back. If you decline microphone access, the coach works as a text conversation instead.

Transcripts and coaching summaries: nothing you say or type in a coaching session is kept unless you switch on "keep a transcript" for that session. It is off by default. If you switch it off again, the transcript and any summary generated from it are deleted. You can also delete an individual session, and deleting a briefing removes its audio and every coaching session attached to it.

We do not use your rehearsals, briefings, or speaker profile to train AI models.

Daily briefs: the assistant checks your gigs and inquiries on a schedule and writes short notes on what needs attention. It sees only your own records to do this. You can dismiss any brief, and the notes are deleted with your account.

Daily brief emails: when there is something worth telling you, we email that brief to your account address through our email provider (Resend). The email contains only your own hub data. Every one has a one-click unsubscribe that needs no sign-in, and turning it off stops the emails without changing anything else in your account. Your briefs still appear on your hub either way.

Subscribing to the blog: if you enter your email address on the blog, we store that address, a normalized copy of it used to stop the same inbox being signed up twice, a one-way hash of the IP address the request came from, and the dates you asked and confirmed. You do not need an account, and we do not ask for a name.

Nothing is sent to that address until you click the link in the confirmation email. If you never click it, the pending record is all that exists and it is never mailed again.

What you get: one email per new blog post, sent through our email provider (Resend). No other email is sent to a blog address, and the list is not used for anything else. Every post email has an unsubscribe link that needs no sign-in and works in one click, including from your mail client’s own unsubscribe button.

To have your address removed from our records entirely rather than marked unsubscribed, contact us at the address at the end of this policy.

You can connect Speaker Hub to an AI app you already pay for, such as ChatGPT, Claude, Claude Code, or Cursor. This works over the Model Context Protocol, an open standard those apps use to reach outside services. It is off until you turn it on.

This is a different kind of sharing from everything else on this page, and it is worth being plain about. Everywhere else, we send your data to an AI provider under our contract with them, and we can tell you what that contract says. Here, you are instructing us to hand your data to an app running on your account, under your agreement with that provider. Once it arrives there, what happens to it, how long it is kept, and whether it is used for training are governed by your agreement with that provider, not by this policy. It is closer to exporting your data to a tool you chose than to us hiring a subprocessor.

What a connected app can reach depends on your plan and is listed on the approval screen before you approve anything. At most it covers your gigs, booking inquiries and their notes, keynotes, coaching tips, social posts, public one-sheet, speech analyses, and the full text of talks you have analyzed, plus the ability to add or update gigs and inquiries on your behalf.

What a connected app can never reach: your email, your calendar outside Speaker Hub, your password, your billing details, and any other speaker’s data. The connector does not expose our Gmail or Calendar tools at all.

Connection records: when you approve an app we store which app it is, when you approved it, what it was allowed to do, and when it last used the connection. We store access credentials only as a one-way hash, so they cannot be read back out of our database. You can see every connected app and disconnect it at any time on the Connect page, and disconnecting takes effect on that app’s very next request.

Personal access tokens: if you create one for an app that cannot sign in through a browser, it is shown to you once and stored only as a hash. Treat it like a password. You can revoke it at any time on the same page.

We do not use anything a connected app reads to train AI models, and we do not receive a copy of your conversations with that app.

If you connect a social account (such as X, YouTube, TikTok, Instagram, Facebook, or LinkedIn) to publish clips, we receive and securely store the access tokens those platforms issue so we can post on your behalf at your request. We request only the permissions needed to publish the content you choose.

You can disconnect a linked account at any time, which revokes our stored access. Each connected platform also handles your data under its own privacy policy and terms.

To make the tools more useful, we may remember preferences and details you share across the app (such as your goals, speaking topics, and your website) so that features stay consistent and tailored to you. This memory powers personalized results across the app, including the SEO tips we generate for your site. You can view, change, or clear this information, or turn memory off entirely, from your account settings.

If you publish a one-sheet, it's hosted at a public web address (for example speaker-hub.com/s/your-name) that anyone with the link can open. That's the whole point of it, so event organizers can find and book you. Because the page is public, search engines and AI answer engines can crawl, index, and cite it, and we include published one-sheets in our sitemap to help them find it. Only the details you choose to put on your one-sheet appear there.

Images you add to your one-sheet, such as your headshot, stage photos, and client or press logos, are stored at public web addresses so the page and its printable version can display them to anyone you send the link to. A direct link to one of those image files keeps working on its own, separately from the page it sits on.

Having a public one-sheet is your choice. You can change or remove it at any time from your settings; once it's removed the page stops loading, and search engines drop it from their results over time. Deleting your account removes the images you uploaded for it.

After the Talk gives a speaker a link, usually shown as a code on their final slide, that anyone in their audience can open. Nobody needs an account to use it. If you fill it in, we store what you send: a rating out of five, what you say stuck with you, and, only if you choose to add them, your name and email address and a note about your own event.

Two of the choices on that form decide what happens next, and both are off unless you turn them on. Ticking that the speaker may quote you publicly is what allows what you wrote to appear on their public speaker page. Ticking that you would like them to speak at your event sends your name, email, and note to that speaker's booking inbox so they can reply to you.

What you send goes to that one speaker. We do not sell it, use it to build a profile of you, or add you to any mailing list of ours. The speaker can read, keep, and delete it, and can contact you if you gave them a way to. To have your response removed, ask the speaker, or contact us at the address at the end of this policy and we will remove it.

If you open an invite link, we store the invite code in a cookie in your browser for up to thirty days so we can credit the member who invited you. The cookie holds the code and nothing else, and it is cleared once your account has been linked or the code has been refused.

When you create an account through an invite link, we record which account invited you, the code used, and a one way hash of the network address your account was linked from. We never store the address itself. The hash is used to tell whether the invite and the new account came from the same connection, which is how we identify people rewarding themselves through second accounts, and it is compared against the same signup records described under security below.

The member who invited you can see that an account signed up through their link and whether it went on to start a trial. They cannot see your name, your email address, or anything you do in the product.

We use strictly necessary cookies to keep you signed in and secure your session, and we store a few functional items in your browser (such as your cookie choice and your most recent tips) to make the tools easier to use. We do not use third-party advertising or cross-site tracking cookies.

You can choose Necessary only or Allow analytics on your first visit. Allow analytics creates random first party analytics and session identifiers. You can withdraw permission and stop future optional collection. For full details, see our Cookie Policy.

We share information only with the service providers that make the product work: for example authentication and database hosting (Supabase), AI processing (OpenAI and Anthropic, plus the open Speech Analyzer models from Meta and NVIDIA, which run on Modal GPU infrastructure), payments (Stripe), email delivery (Resend), and any social platforms you choose to connect (to publish the content you ask us to), and only to the extent needed to provide the service.

Separately from those providers, if you connect Speaker Hub to your own AI app, your hub data goes to that app on your account and under your agreement with its provider. See "Connecting Speaker Hub to your own AI" above, because that sharing works differently from everything in this section.

We may disclose information if required by law, to protect our rights, or as part of a business transfer such as a merger or acquisition. We do not sell your personal information.

If you choose to make a voluntary donation to support Speaker Hub, your payment is processed by Stripe, the same processor we use for subscriptions. We receive confirmation of the donation and limited details (such as the amount and, if you provide it, your email), but we never see or store your full card number.

If you donate through a third-party platform we link to, that platform collects and processes your payment under its own privacy policy and terms. Donations are optional and are never required to use the Service.

We keep your information for as long as your account is active or as needed to provide the service. You can delete your content, and you may request deletion of your account and associated data.

Optional product analytics events are deleted after 13 months. Marketing leads are kept only while needed for the requested follow up or consented product communication. Withdrawing optional email permission suppresses queued lifecycle messages immediately.

ClipFlow source videos, whether you uploaded them or we fetched them from a YouTube link you gave us, and their private thumbnails are automatically deleted after 30 days. Rendered clips remain in your account until you delete the ClipFlow project or your account. Deleting a ClipFlow project removes its stored source, thumbnail, rendered clips, transcript, and clip records immediately.

A QuickCut project keeps the talk you uploaded, its transcript, and any reel you rendered until you delete the project or your account. Deleting a QuickCut project removes its stored video and its rendered reel immediately. A QuickCut project linked to YouTube has no stored video to remove.

Speaker Hub no longer runs a Chief of Staff chat or accepts new chat conversations. We do not receive a copy of your conversation in ChatGPT, Claude, or another connected app, and we do not store provider API keys, chat voice notes, or chat attachments. Empty legacy chat tables are retained temporarily during a decommissioning grace period and will be removed in a later release; account deletion still covers any associated legacy data.

Some things are deliberately not kept at all: raw microphone audio from a coaching session is never stored, and a coaching transcript or summary exists only if you turned that on for that session. Deleting an Audience Insider briefing also deletes its generated audio and every coaching session attached to it.

We use industry-standard safeguards, including encryption in transit, to protect your information. No method of transmission or storage is completely secure, but we work to protect your data and limit access to it.

Depending on where you live, you may have the right to access, correct, export, or delete your personal information, and to object to or restrict certain processing.

To exercise these rights, or if you have any questions about this policy, contact us at mick.hatzikostas@gmail.com.

We may update this policy from time to time. When we make material changes, we will update the date at the top of this page and, where appropriate, notify you in the app or by email.