Effective 17 September 2026 · Version 1.0
This policy explains what Mentara Persona (the “Service”) collects, why, who it is shared with, and how to get it deleted. It forms part of our Terms of Service.
Your email address, authentication credentials managed by our authentication provider, and your plan, credit balance and purchase history.
The Service captures audio from the device you select. Audio is processed to produce a transcript. Audio segments and the corresponding transcripts are written to log files on your own computer so that the Service can be debugged and improved.
We do not create voiceprints or speaker embeddings, and we do not perform speaker identification. We do not sell, licence or distribute audio recordings.
Text produced from the audio, and the text of the persona's replies.
This is the bulk of what we keep and the part that improves the Service. It includes: speech-recognition confidence scores; timing measurements for each stage (recognition, model, speech synthesis, total perceived latency); whether speech was interrupted or talked over; whether the persona chose to respond or stay silent and why; how long a turn was held before responding; which persona and mode were active; and moderation events.
None of this requires knowing who you are.
Session start and end, duration, and credits consumed.
When you visit mentarapersona.com we record which pages you view and which links you click (Discord, download), along with where you came from: the link or site that sent you, such as a video or a partner's link. Your browser keeps a random ID so repeat visits count once, and remembers where you came from for 90 days. We don't record your IP address, and this doesn't identify you. If you create an account, we note where the account came from, and in the app you can tell us how you heard about us if you like.
We use interaction data to develop, train, fine-tune and evaluate machine learning models, and to create derived datasets.
A derived dataset is built from the measurements of a conversation rather than its content. A record consists of things like: how long the speech was, the recognition confidence, the gap before the person spoke again, whether the persona responded, how long it took, and the label for what decision was taken. Content fields — the words spoken, the reply, the audio file — are removed before any dataset leaves our control.
These derived datasets may be licensed or distributed to third parties. Audio recordings, voiceprints and speaker embeddings never are.
Where required by law, we will obtain your consent for this use, and you may withdraw it. Withdrawal applies going forward; it does not require us to retrain models already trained or to recall datasets already distributed.
The Service cannot function without sending data to specialised providers. We do not sell personal information to advertisers or data brokers.
| Provider | What it receives | Why |
|---|---|---|
| Deepgram | Audio | Speech recognition |
| ElevenLabs | Reply text | Speech synthesis |
| Cerebras, Groq, OpenAI, Anthropic | Conversation text | Response generation (whichever is active or serving as fallback) |
| Supabase | Account, billing, usage records | Authentication and account data |
| Stripe | Payment details | Payments, when purchasing credits |
| Vercel | Website requests | Website hosting |
We may also disclose information where required by law, to enforce our Terms, or to protect the rights and safety of people.
/forget command
clears what the persona remembers about a conversation.We will not discriminate against you for exercising these rights.
If you use the Service in a conversation involving anyone else, their speech is processed too. You are responsible for telling them and for obtaining any consent the law requires — see Section 6 of the Terms.
We treat a third party's data as necessary to operate the Service for you. We do not use speech from people who have not accepted these terms to train models or to build derived datasets, and our pipeline separates the two cases so this can be enforced rather than merely promised.
If you believe your voice was processed by the Service without your agreement, contact us and we will delete what we hold.
Credentials are stored using the operating system's protected storage on your machine, and communication with our services uses encrypted connections. No system is perfectly secure; we cannot guarantee absolute security.
The Service is for adults aged 18 and over. We do not knowingly collect data from children. If we learn that we have, we will delete it.
Our providers operate in the United States and elsewhere. Using the Service involves transferring your data internationally, including to countries whose data protection laws differ from those where you live.
We may update this policy. Material changes will be announced before they take effect, and each version carries a version number and effective date so it is clear which applied when.