Voice by default in all demos

We're upgrading all text demos to voice conversations by default. We've been running voice demos for a few months now, and on average they are more engaging than their text counterparts.

What made this move possible is that voice now supports every language we support for demos. Whatever language a demo runs in, the agent can speak it, and visitors who arrive in a language we don't support get their demo in English.

Everyone can still choose to type. A visitor who declines the microphone prompt can write their replies in the chat panel, which opens automatically, or tap one of the suggested replies the agent offers after each turn. The agent won't ask them to say anything out loud.

Visitors can also change their mind in the middle of a demo. Enabling the microphone later upgrades the conversation to voice on the spot, in the same session, without restarting the demo or losing what has already been said.

In agent settings you can choose a voice for English, German, French and Spanish, and play a sample in any of the seven supported languages. Estonian, Latvian and Lithuanian currently use fixed voices.

If you embed demos through the SDK, nothing changes on your side. The modality option on openDemo and on embed URLs is simply ignored now, and existing embeds keep working as they are.

Inbound Q&A and Onboarding agents are unaffected. See supported languages for the languages available to each agent type.