Privacy notice

Last updated 2026-07-04

Heads-up - closed tester period (since May 2026)

Important. Command Voice is in a closed-tester period with a small group of invited testers. During this window - until the notice below is updated to say otherwise - your onboarding voice sample, each original audio recording, and the voice-clone reply generated from each take are retained on our server so the operator can review them for product effectiveness. This is in addition to the normal client-side storage described below.

By recording, you consent to this retention. If you don't want your audio retained server-side, stop recording until this notice is updated, or email the operator to opt out and we'll purge what we hold from you within 24 hours.

What this notice is - and isn't

A plain-language description of what Command Voice collects, where it goes, how long it stays, and how you can have it deleted. It's accurate as of the date above and the source of truth for our privacy posture in the current early version.

When we commercialise we'll replace this with a fuller policy covering Information Officer appointment, data-flow diagrams, and processor contracts. Until then, this page reflects exactly what the code does.

Who runs Command Voice

Built by Mali (founder of The Crafted Persona) and a small group of early users. Reach the operator at mmkela@gmail.com.

What we collect

Your account - Email address, password stored as a hash, your cell number (used for practice reminders on WhatsApp), coach-visibility consent timestamps, whether and when you connected WhatsApp reminders, and your opt-out if you reply STOP.

Your name (optional) - From the onboarding form or your account. Used to greet you on the home screen and label your rows in operator views.

Voice recordings - The audio you record during course lessons, deck rehearsals, or live analysis. Stored in your browser; transmitted to third-party services for analysis (see "Where it goes").

A voice sample of about three minutes - Used to create a voice clone so corrections can be played back in your voice. The clone lives with our voice-clone service. During the closed-tester window, a copy of the original onboarding sample is also kept on our server for operator review.

Transcripts, metrics, corrections, coaching text + audio - Generated by our analysis pipeline from your recordings. Cached in your browser so they don't have to be regenerated.

Course progress - Which lessons you've completed and the days you practised. Kept in your browser and mirrored to our server so the operator can see how the tester group is progressing. No audio or transcripts travel with it.

If a friend invited you - The invite code from their link is stored with your account so the operator can see who invited whom. Nothing else connects you to them.

Your account id - Sent to our server with paid calls so we can attribute costs. Some pre-account browser history may carry an older local id.

Feedback you send - The feedback type, description, optional contact email, page URL, browser details, viewport, timezone, and any optional screenshot you attach from the in-app feedback widget.

Where it goes

Your browser, on your device - The primary store for your profile, voice sample, course progress, every recording, transcript, correction, coaching audio, and metric cache — held in your browser's own storage. Stays here unless you wipe it via /account.

Our server (service-cost log) - One line per paid analysis call: the service used, duration, computed cost, your account or local id, the session and slide it relates to. No audio or transcript content.

Our server (progress mirror) - The lesson numbers you've completed and your practice days, so the operator's progress view works. Nothing you said is included.

Our server (closed-tester audio capture) - During the closed-tester period - see the amber notice at the top of this page - the onboarding voice sample, each original recording, and each voice-clone reply are also written to a private operator directory, keyed by your account or local id and session id. Off by default outside this window. Operator review only, behind the operator's sign-in.

Our server (feedback log) - Feedback messages are written to a private JSONL log. Optional screenshots are saved beside that log and may also be summarized by email to the operator.

Speech-to-text service (US-based) - Raw audio is uploaded to a third-party speech-to-text service for transcription. Their retention follows their standard contract; the full list of third-party services we use is available on request from the operator email above.

Voice-tone analysis (on your device) - How steady and lively your voice sounds is measured on your device, inside your browser. Since June 2026 no audio leaves your browser for tone analysis.

WhatsApp reminders (only if you connect them) - You start the connection by messaging our WhatsApp number. Your cell number and reminder messages travel through WhatsApp and the messaging service we send from; each reminder sent and its delivery status is logged so the operator can see what went out. Reply STOP at any time to stop them.

Voice-clone service (US-based) - Your voice sample is uploaded to a third-party voice-cloning service. The resulting clone is kept under a permanent voice id until you ask us to delete it (/account - Delete voice clone).

Large-language-model service (US-based) - Transcripts are sent to a third-party language model to generate corrections and coaching text. No audio, no profile content.

How long we keep it

In your browser - Indefinitely, until you wipe it via /account or the browser clears its storage. Automatic age-based retention lands before commercialisation.

On our server (service-cost log) - Entries persist for the life of the closed-tester period as the operating cost record.

On our server (closed-tester audio) - Retained for the duration of the closed-tester window, and an automatic weekly cleanup deletes any stored audio older than 90 days regardless. Everything is purged when the operator switches the capture flag off.

On our server (reminder log) - Each WhatsApp reminder sent and its delivery status, kept as the operating record. Email the operator to have your rows purged.

On our server (feedback log) - Kept while we use it to investigate issues and improve the closed tester build. Email the operator to have your feedback rows and screenshots purged.

With third-party services - Their standard retention applies. Voice clones are permanent until explicitly deleted.

Your rights (POPIA + GDPR equivalents)

Access - Download everything this browser holds via /export. For server-side records, email the operator and we'll return your slice within 7 days.

Deletion - Wipe everything in this browser via /account. Delete your voice clone via the same page. For server-side records, email the operator and we'll purge your rows within 7 days.

Correction - Edit your profile via /onboarding, or your name, email, and cell number via /account.

Withdrawal of consent - Wipe your browser data + delete the voice clone. Reply STOP to any WhatsApp reminder to stop those. Going forward your continued use of the product is your continued consent.

Cross-border data transfer

Audio you record leaves South Africa and is processed by third-party services based in the United States. They've represented standard contractual safeguards. Full Data Processing Agreements with each service are in place before commercial launch. If you connect WhatsApp reminders, those messages also travel through WhatsApp, which has its own privacy policy.

Changes

We update this notice whenever the underlying data picture changes.