Voice

What voice can and cannot do in Stepora today.

Voice-note replies — available now

The bot can answer with voice notes, using free built-in text-to-speech voices. There are 54 preset voices to pick from, and no extra charge for them — the voice itself costs $0, so voice-note replies are free. Voice-note replies are included on Growth and Business plans; Starter is text chat only.

Languages that work:

• English (US and British voices)
• French
• Hindi (Devanagari script)
• Roman Urdu (read with English phonemes — audio is produced, quality varies)

Pick the voice and language in the panel. If a voice and language do not match, the panel refuses with an error instead of producing the wrong audio.

Voice cloning — not available

Cloning a voice from a sample is not available. The cloning weights need an access approval that has not been granted yet, and one provider (Kokoro) cannot clone at all. Until that changes, preset voices are the only option. Do not send voice samples expecting a clone — they will not be used.

Live voice calls — not yet available

Live voice calling (you or the AI agent talking to a customer on a WhatsApp call) is not yet available. When it arrives, AI voice agent calls will be priced at $0.05 per minute. Meta's own calling minutes are separate: inbound calls are free from Meta, outbound calls are charged by Meta per minute by destination country.

Speed, honestly

Voice notes are generated as whole audio files, not streamed. On ordinary server hardware that runs at roughly real-time speed — a 15-second reply takes around 15 seconds to generate. That is fine for voice-note replies. It is one reason live calling is not offered yet.