Voci listens to you speak and remembers what you have learned. This page says exactly what that means in terms of data — what is kept, what is sent elsewhere, and how to get rid of it.
| Data | Why |
|---|---|
| Email address and name | To identify your account and send account-related mail. Your password is stored only in hashed form — it cannot be read back, including by us. |
| Your CEFR level | Set by the placement test. It decides which words are worth teaching you. |
| Review history | Which words you have seen, how you graded, the scheduling intervals, and when each word reached maturity. This is what the spaced-repetition schedule and the dated forecast are computed from. |
| Pronunciation scores | For each attempt: the sentence, an overall score, which words were weak, and a per-sound score list. The recording itself is not stored for these attempts — only the numbers that came back. |
| Benchmark voice recordings | If you use the before/after feature, Voci keeps dated audio of you reading one fixed sentence, so you can hear your own change over months. These are stored, capped in number, and deletable by you at any time. |
| Voice twin reference and generated clips | On the Pro+ Voice plan a synthetic copy of your voice is created by our voice-synthesis provider. Voci stores its identifier and the synthetic clips generated for you, so replay does not send the text or spend your allowance again. |
| Subtitle files you add | The dialogue text is extracted to build the word list for that film. |
| Subscription state | Which plan you are on, whether payment is healthy, when the period ends, and a customer identifier from the payment provider. Voci never sees or stores your card details. |
There is no advertising on Voci, no advertising identifiers, no third-party analytics or tracking scripts, and no sale of personal data to anyone, ever.
Voice data gets its own section because it deserves one. A recording of your speech is more personal than an email address, and Voci's whole method depends on hearing you.
When you record a word or a line, the audio is sent to our speech-assessment provider, which returns an accuracy score and a score for each individual sound. Voci keeps those numbers and discards the audio. The provider processes it to produce the result and does not use it to train models. They are named in section 3.
These are the exception: the audio is kept, because the feature is worthless without it — hearing week 1 against week 12 requires week 1 to still exist. They are stored on Voci's own servers in Canada, never sent to a third party, limited to a small fixed number of takes, and erased when you delete them. Deletion removes them outright rather than hiding them.
On Pro+ Voice, recordings you provide are sent to our voice-synthesis provider to create a synthetic voice that can read any line in your own voice. This is voice cloning, and it happens only when you deliberately set it up. You can ask us to delete the cloned voice at any time and it is removed from the provider as well as from Voci.
Audio produced by your voice twin is synthetic — generated by software, not recorded from you. Generated clips are kept on Voci’s servers in Canada for your private replay, so playing or slowing the same line does not generate or send it again. They are deleted with the twin or your account.
A voiceprint is biometric data. Some laws treat it as a special category with its own rules, and this section is written to meet them rather than to satisfy them by accident.
What it is. The voice twin, and the benchmark recordings you keep for the before/after feature. Nothing else. Pronunciation scoring produces numbers and discards the audio, so it creates no biometric record.
Why we collect it. One purpose only: to let you hear a line, or your own progress, in your own voice. It is never used to identify you, never used to train anyone's models, never sold, leased, traded, or otherwise profited from, and never shared with anyone outside the providers named in section 3.
How long we keep it. Whichever of these comes first:
That last one is a hard ceiling and it runs automatically. Destruction means the twin is deleted at the provider and the reference removed from Voci; recordings are erased from our servers rather than hidden or archived.
Only these, only for the stated purpose, and none of them for advertising. This is the single place providers are named, so that it stays accurate — if one changes, this table changes and the date below changes with it.
| Provider | What they receive, and why |
|---|---|
| Paddle.com Market Ltd payments (web) | Merchant of record for web subscriptions. They receive your email and payment details directly and handle the transaction, tax and invoicing. Voci receives back only a customer id and the subscription status. |
| Apple payments (iOS) | For subscriptions bought in the iOS app. Apple handles the payment; Voci receives a signed transaction which it verifies. |
| Microsoft Azure speech assessment | Audio of the phrase you just spoke, to score its pronunciation. Returned as numbers; the audio is not kept. |
| ElevenLabs voice synthesis | Only if you enable the voice twin: the recordings used to build it, and the text of lines you play back. |
| Anthropic scene writing | The words you are learning, to compose the weekly practice scene. No account details are sent. |
| Google sign-in, hosting | If you choose "Sign in with Google": your email and name. Google Cloud also hosts the service, in Canada. |
| OpenSubtitles, TheTVDB film data | Film titles and subtitle lookups. These are queries about films, not about you. |
Providers in this table last reviewed 31 July 2026. If one is replaced, this page is updated and account holders are told when the change is material.
Voci will also disclose data if legally compelled to, and will tell you unless prohibited from doing so.
Your account and learning data are stored in Canada. Speech scoring and voice synthesis are performed by the providers in section 3, which may process it in another country.
Data is kept while your account exists. Delete your account and the account row, review history, pronunciation scores, benchmark recordings and voice twin reference go with it. Billing records are retained by Paddle for the period their own tax obligations require, which is outside Voci's control.
Wherever you live, Voci will honour the following. If you are in the EU/UK these are your GDPR rights; in Canada, your rights under PIPEDA.
Email support@voci.ca and you will get a reply within 30 days. There is no charge.
Voci sets no tracking or advertising cookies. Your login session is held in your browser's own local storage so that you stay signed in; clearing your browser data signs you out. Paddle sets its own cookies inside its checkout, governed by Paddle's privacy policy.
Voci is for people aged 16 and over, or 13 and over where local law lets someone that age consent to their data being processed and a parent or guardian agrees. Accounts are not knowingly created for anyone younger. If you believe a child has registered, email support@voci.ca and the account, including any voice data, will be removed.
Traffic is encrypted in transit. Passwords are stored hashed, never in a form anyone can read. Payment card data never reaches Voci's servers at all. No system is perfectly secure, and if a breach ever affects your data you will be told promptly and plainly rather than in a footnote.
If this policy changes materially, the date at the top changes and account holders are emailed. Continuing to use Voci after that means accepting the revision.
Mohammad Parvizi, trading as Voci · Ontario, Canada
support@voci.ca