Voice and read-aloud, comxbot help

What voice does

Every assistant can read its answers aloud, so a visitor can listen rather than read, and they can talk to it instead of typing. This helps people who find reading hard, anyone using a phone with their hands full, and learners practising spoken English.

Read-aloud works out of the box using the voice built into the visitor's browser, at no extra cost and with nothing to set up. It's on by default.

Bringing a studio-quality voice

If you want a natural, human-sounding voice, you bring your own key from a text-to-speech provider: ElevenLabs, OpenAI, Azure, Google or PlayHT. The audio bills to your account with that provider, the same bring-your-own-key model as the chat AI.

You set the key for the whole workspace or for a single assistant, and keys are encrypted at rest. If no studio key is set, the assistant simply falls back to the free browser voice, so read-aloud always works.

Different voices per assistant and language

You can give each assistant its own voice, and map a different voice to each language, so a reply in Spanish is read by a Spanish voice and a reply in English by your chosen English one. This matters for multilingual and ESOL audiences.

Audio streams as it's generated so playback starts quickly rather than waiting for the whole reply, which keeps spoken answers feeling responsive.

Keeping costs under control

Studio voice usage is metered, and you can set a daily spend cap per workspace so a busy day, or misuse, can never run up an unexpected bill. When the cap is reached the assistant falls back to the free browser voice rather than stopping.

There's a live demo on the voice page where you can compare the browser voice with a studio voice before you decide.