Technology

Voice cloning, done responsibly

Turn a clear voice recording into a private synthetic voice for your companion. Famyl only clones voices with consent or legal authority.

Voice cloning, done responsibly

Turn a clear voice recording into a private synthetic voice for your companion. Only clone your own voice or a voice you have permission and legal authority to use.

How it works

Upload a clear recording. Famyl learns the voice’s tone, accent and rhythm so your companion’s replies can be heard in that voice.

Consent and privacy

  • No celebrities, public figures or religious figures
  • No living person’s voice without their consent
  • Voices are never sold or shared
  • You can delete a preserved voice at any time

What to record

Use 1–3 minutes of clear speech from one speaker, with as little background noise as possible.

  • Record in the app: at least 20 seconds and up to 3 minutes, in a quiet room with the phone about a hand's width away
  • Or upload what you already have: MP3, M4A, WAV, OGG/Opus voice messages or a standard phone video (up to 200 MB and 30 minutes)
  • Uploads are prepared on your own device as a clean mono copy; the original file is kept unchanged and private
  • Natural conversation works best — for example, ask “What was your first job?” or “Tell me about the house you grew up in”
  • Only upload a voice you have the right to use; for a living person, get their permission first

Voice cloning languages

  • Arabic
  • Bengali
  • Bulgarian
  • Chinese
  • Croatian
  • Czech
  • Danish
  • Dutch
  • English
  • Filipino/Tagalog
  • Finnish
  • French
  • Georgian
  • German
  • Greek
  • Gujarati
  • Hebrew
  • Hindi
  • Hungarian
  • Indonesian
  • Italian
  • Japanese
  • Kannada
  • Korean
  • Malay
  • Malayalam
  • Marathi
  • Norwegian
  • Odia
  • Polish
  • Portuguese
  • Punjabi
  • Romanian
  • Russian
  • Slovak
  • Spanish
  • Swedish
  • Tamil
  • Telugu
  • Thai
  • Turkish
  • Ukrainian
  • Urdu
  • Vietnamese

Frequently asked questions

Can AI preserve someone's voice?
Yes. A few minutes of clear recorded speech is usually enough to build a voice model that reproduces someone's tone, accent and rhythm convincingly.
Is voice cloning legal?
Cloning your own voice, or a deceased relative's voice with family authority, is generally lawful in most jurisdictions. Cloning a living person without consent, or any public figure, is not — and Famyl blocks it.
How much audio does voice cloning need?
Typically one to three minutes of clean speech for a good result, and more for the highest fidelity. Background noise matters more than length.
Which languages can Famyl clone a voice in?
You can choose from 44 languages for a cloned voice, including Arabic, English, Hindi, Japanese, Odia, Spanish and Urdu. Search the full list at famyl.ai/voice-cloning#voice-languages. Persian isn't available yet. Accent, dialect and audio quality can vary. This is separate from the website's display languages, from free text chat, and from Isaac Newton's Live voice, which speaks British English only.
Can I delete a cloned voice?
Yes. Preserved voices can be deleted from your account at any time, and deletion removes the model as well as the source audio.

Related