AI Voice Policy

Last updated 23 October 2026

This page explains how our voices are made, what we will never do, and what we expect from you when you publish AI-generated speech.

No voice cloning. Cylindricle does not clone anyone's voice and does not accept voice samples or recordings. Every voice is an original synthetic voice with an invented name.

1. Where our voices come from

Cylindricle uses the open-source Kokoro text-to-speech model (Apache 2.0 licence), whose published documentation describes its training data as permissively licensed or non-copyrighted audio and synthetic data. We rely on the model's publisher for that description and cannot independently verify it. Our named voices (such as Aria or Atlas) are voice styles that ship with the model. We do not present any of them as, and have not built any to imitate, a real person, celebrity, politician or other public figure.

2. What we do not do

We do not offer voice cloning, voice conversion or "sound like this person" features.
We do not accept uploads of voice recordings.
We do not name or market any voice as a real person's voice.
We do not use your text or audio to train AI models.

If we ever add a cloning feature in the future, it will only ever work with a person's own voice, with verified, recorded consent from that person, a clear opt-in and revocation process, and an update to this policy before launch.

3. AI labelling

Every MP3 and WAV file we produce carries a metadata comment stating that it is AI-generated speech made with Cylindricle. Please do not strip it.
We support labelling requirements such as those in India's IT rules on synthetically generated content, and we will add further labelling features (for example an optional audible or on-screen notice) as rules and standards develop.

4. What you must do when you publish

Do not present AI speech as a real person speaking when that would mislead listeners, for example in news, testimonials, reviews, political or public-interest content, or calls to customers.
Disclose AI narration where the law, a platform's rules (such as YouTube's altered or synthetic content disclosure) or good faith requires it.
Do not use a voice to impersonate a real person, a company, a government body or a bank, or to write fake quotes in anyone's name.
Do not use the audio for fraud, scams, harassment, election interference or non-consensual intimate content.

The rules in the acceptable use policy apply in full to AI-generated speech.

5. Look-alike voices

If you think one of our voices sounds like you or someone you know, tell us at support@cylindricle.com. Our voices are generic synthetic ones, but we will review every report in good faith and retire or change a voice where a real similarity is found.

6. Reporting misuse and takedowns

Report suspected misuse, deepfakes or impersonation to support@cylindricle.com. We acknowledge within 24 hours. For urgent harms, such as non-consensual intimate content or child safety, we act as fast as we can and within the timelines Indian law sets. We remove infringing content, suspend accounts, preserve evidence, and cooperate with law enforcement where required.

7. Limits of the technology

AI speech can mispronounce, misread numbers or names, or stress the wrong word. You are responsible for checking audio before you rely on or publish it, especially for medical, legal, safety or financial information.

Questions about this page? Write to support@cylindricle.com. See also our terms, privacy policy, refund policy and AI voice policy.