Skip to content

Voice

Talk to an agent instead of typing, and let it talk back.

Speaking to an agent

The composer's microphone control records what you say and turns it into text. Review the transcript like any typed message — nothing is sent until you press Enter, so you always get a chance to fix a misheard word first.

Hearing an agent's replies

Each agent reply can offer a speaker control that reads the message aloud. Whether it's there depends on how the agent is configured — an agent with no voice set stays silent, with no speaker button shown.

Choosing your voice engine

By default, playback and dictation use your browser's built-in speech engine — no setup, and no audio ever leaves your device.

If your organization turns on the platform's self-hosted voice engine (Voicebox, on Organization → Voice, admins only), everyone gets a second option. Open your voice preferences on an agent's settings page and pick an Engine:

  • Browser — your device's own speech synthesis and dictation. Nothing leaves your machine.
  • Voicebox (local) — a higher-quality, cloned-voice engine that runs on our infrastructure rather than a third-party voice vendor. Pick one of your organization's configured voice profiles and a playback speed.

Your choice is personal — it doesn't change what anyone else on your team hears. If an admin turns Voicebox off for the organization, everyone reverts to the browser engine automatically; your saved preference is kept in case it's turned back on later.

Voicebox is opt-in

Voicebox has to be enabled for your organization before it shows up as a choice. Ask an owner or admin to turn it on from Organization → Voice if you don't see it in your preferences.

Privacy

With the Browser engine, dictation and playback never leave your device — there's no clip to upload, so nothing is transmitted or stored.

With Voicebox, your recorded clip is sent to the platform to be transcribed, and — like any uploaded file — it's scanned for malware first. Either way, audio is only ever used to do what you asked: transcribe it or speak a reply.

On mobile

The mobile app uses your phone's native speech features today; the in-platform Voicebox engine is currently a web-app feature.

See also