Your voice. Their language.

1. Record a transcript. Open the agent builder, hit Clone my voice, and read the script we provide. About 60 seconds is enough; longer samples produce a closer match.
2. We process it. The sample becomes a reference voice attached to your account. No extra training step on your side — the same audio drives both phone calls and the chat widget.
3. Pick it on any agent. Each agent has a voice setting; choose your cloned voice and a language. You can run different languages on different agents (e.g. one agent for English-speaking patients, another for Spanish), or one agent that switches based on the caller.

• Read in a normal speaking voice — the phone tone you'd use with a customer, not announcer-mode.
• Quiet room, no echo. A USB headset mic is more than enough; you don't need a studio.
• One person, one sample. The clone reproduces a single speaker's timbre.
• Sample length: 30 seconds works, 60 seconds is the sweet spot, beyond two minutes adds little.

The voice itself stays the same; the agent picks the language based on either:

• The agent's configured language (set in the agent builder), or
• What the caller is actually speaking — if you enable language auto-detect, the agent matches the caller turn-by-turn.

Auto-detect is useful for clinics and salons whose customers code- switch (English ↔ Spanish, English ↔ Hindi). Stick with a fixed language if you want predictable behaviour for a single audience.

Every language below works on the chat widget and on phone calls — same model, same voice. ISO 639-1 codes shown for the few cases where they matter (API params, agent config).

The ReasonVoice TTS layer normalises numbers, times, dates, and currency before speaking — so “$1,250.00” comes out as “one thousand two hundred fifty dollars” and “3:30 PM” comes out as “three thirty PM” across every language. You don't need to spell anything out in your prompts.

• One cloned voice per account on Starter, up to 5 on Pro.
• Voices are tied to your account, not shared across customers. Your competitor can't pick your voice from a public list.
• Sample uploads accept WAV, MP3, FLAC, OGG. Max 25 MB per file.