Privacy Policy

Account information. If you sign in, we collect the email address you sign in with. By default, SpeakPrint uses passwordless, one-time email links for sign-in. If you choose to set a password in Settings for faster sign-in, Supabase stores it hashed — SpeakPrint never sees or stores it in plain text.

Recordings and transcripts. When you record or upload audio, that audio is sent to our transcription provider to generate a timed transcript (see "Who else processes your data" below). If you're signed in, the audio and transcript are then saved to your account so you can return to them later. If you're not signed in, the upload is held in private temporary storage for recovery. Access expires 24 hours after the upload is created; scheduled cleanup then removes the storage object and job row, while successful cancellation removes them sooner. To prevent accidental loss during a failed request, SpeakPrint also keeps one temporary recovery draft in this browser storage. The draft can include the audio, transcript, and practice focus. It expires 24 hours after capture and is removed the next time SpeakPrint opens; a successful account save or starting over removes it sooner. It is never shared with another website.

Practice-session context. If you're signed in, SpeakPrint saves the title, selected practice purpose, optional target date, and optional goal you provide so related attempts can stay together and be compared. This context stays with your account and follows the same deletion choices as your recordings.

Abuse prevention. When a transcription is requested, SpeakPrint creates a keyed, one-way identifier from either your account ID or network address to enforce request limits. The rate-limit table does not store the original account ID or network address.

Derived analysis. SpeakPrint computes relative pace, timing, repetition, transcript-certainty, and optional audio-boundary observations. This experimental analysis is generated on the fly from data described above—it is not a diagnosis or a separate data source.

Product events and optional feedback. SpeakPrint records a limited set of action names and bounded categories—such as whether a signed-in user completed an analysis, began a second attempt, or found a result understandable. It does not place audio, transcripts, titles, goals, recording identifiers, or email addresses in these events. Optional written feedback is limited in length and should not contain private recording content.

Pilot-interest information. If you submit the pilot form, SpeakPrint stores the contact name, email, organization, approximate group size, intended use, optional message, and your explicit consent to be contacted. A pilot-interest submission is not a partnership announcement.

Practice Briefs. If you create a Practice Brief, SpeakPrint stores its title, purpose, instructions, optional target date, group type, and a high-entropy invitation code. If you explicitly join one, SpeakPrint stores an association between that brief, your account, and a participant-owned Practice Session. The brief creator receives only aggregate counts for joined, started, repeat-attempt, and completed sessions. SpeakPrint does not provide them your identity, recording, transcript, Voiceprint, or individual result.

Subscription records, if billing is activated. SpeakPrint may store your current plan state, Stripe customer and subscription references, billing period end, and bounded webhook-processing records. SpeakPrint does not receive or store complete card numbers or payment credentials. Checkout remains disabled unless the public Pricing page and required billing configuration are active.

Plan-usage records, if plan limits are active. SpeakPrint stores an account-scoped analysis timestamp and opaque session reference to enforce lifetime or rolling plan allowances. It does not copy audio or transcript content into this record. Deleting one recording does not reset consumed analysis usage; deleting your account deletes the usage record.

We rely on a small number of infrastructure providers to run SpeakPrint, rather than building and hosting everything ourselves:

• AssemblyAI receives your audio to produce a transcript with word-level timestamps. This happens whether or not you're signed in. SpeakPrint requests deletion of the provider's audio and transcript as soon as the completed result has been received, and does not return the analysis when that cleanup cannot be confirmed.
• Supabase provides authentication, database storage, and file storage for saved recordings, transcripts, and account data.
• Vercel hosts the website and provides aggregate performance analytics. SpeakPrint does not intentionally send recording audio or transcript text to Vercel Analytics.
• Plausible receives aggregate product events, such as whether a recording or practice step was completed. These events use a small allowlist of non-identifying categories and counts. They do not include audio, transcripts, session titles, goals, email addresses, recording IDs, or account IDs.
• Sentry may receive error and performance telemetry when monitoring is enabled. SpeakPrint removes request bodies, cookies, query strings, exception messages, arbitrary context, and direct user identifiers before an event is sent. Only a small allowlist of non-identifying operational fields remains; session replay is disabled.
• Stripeprocesses checkout, payment, subscription, invoice, and billing-portal information only if billing is activated and you choose to purchase a plan. Payment credentials are handled by Stripe, not stored in SpeakPrint's database.

We don't sell your data, and we don't use it for advertising or share it with data brokers.

Your saved recordings, transcripts, and audio are restricted to your own account. Administrator dashboards receive only pseudonymous operational metadata such as recording duration, date, source, and practice focus. They do not provide standing access to transcript text, recording titles, or raw audio. Aggregate founder reporting may include funnel totals, processing duration, bounded feedback responses, and use-case counts. Pilot contact details are accessible only for responding to the request and are not included in aggregate product analytics. A Practice Brief creator receives aggregate milestone counts only; there is no participant roster or creator-facing access to individual speaking content.

Saved recordings, transcripts, and practice-session context stay in your account until you delete them. You can delete an individual recording from your Sessions library at any time, which removes its transcript and audio from our storage. Deleting your account removes the rest. Provider audio and transcript artifacts are requested for deletion immediately after each transcription completes. Unsaved browser recovery drafts expire 24 hours after capture and are removed the next time SpeakPrint opens; a successful save or Start over removes them sooner. Pseudonymous request-limit entries expire after a short abuse-prevention window and are removed by a daily cleanup job. Account-linked product events are retained for up to 400 days. Optional feedback and pilot-interest submissions are retained for up to 18 months unless deletion is requested sooner. Billing webhook event identifiers are retained for up to 400 days for replay protection and operational auditing. Account plan-usage records remain until account deletion so ordinary session deletion cannot bypass an analysis allowance. Practice Brief creators can archive a brief to prevent new joins or permanently delete it from Practice Briefs. Leaving a Practice Brief deletes the association to it while keeping the resulting private Practice Session in your account.

You can download a machine-readable JSON copy of your account email, transcripts, recording and practice-session metadata, quality information, and derived metrics from Settings. Recording audio is deliberately excluded so exporting does not silently create another copy of sensitive audio on your device. You can explicitly download the original audio from an individual saved Voiceprint.

You can delete individual sessions from Sessions, delete your account from Settings, or contact us for help with an access or deletion request. Starting over before saving also removes the current browser recovery draft. Pilot contacts and people who submitted optional written feedback can use the contact address below to request deletion. You can permanently delete a brief you created or leave a joined Practice Brief from Practice Briefs. Aggregate counts update without exposing participant identities.

Data in transit is encrypted (HTTPS). Saved recordings are stored with access rules that restrict each account to its own data. No system is perfectly secure, and as a prototype-stage product, SpeakPrint hasn't undergone a formal third-party security audit yet.

SpeakPrint describes patterns in your recordings — it does not diagnose cluttering, stuttering, or any other condition. Speech recordings can still be personal and sensitive, so SpeakPrint protects them accordingly. Don't use the product as a substitute for professional evaluation.

SpeakPrint isn't directed at children under 13, and we don't knowingly collect information from them.

If this policy changes in a way that meaningfully affects how your data is handled, we'll update the date above and, where practical, note the change on this page.

Questions, deletion requests, or anything else: hello@speakprint.com.