Say what you mean.
Talk with an Agent, dictate a message, or ask it to speak a reply. Streaming captions let you follow what it heard while the conversation develops.
You can interrupt or correct the direction and continue with the same project. A spoken request can lead to a document, a change in an application, or a next step for the team.
Set up voiceBring the discussion into the project.
Meeting connections can bring speech and transcripts into the team’s work. Geyser supports meeting-specific paths for services including Zoom, Google Meet, Microsoft Teams, and Slack Huddles.
Transcription, spoken participation, and customer-hosted processing are configured separately. Your setup determines which meeting actions are available. The meeting provider still handles the media that travels through its service.
Explore connected appsChoose where the voice is processed.
Use approved speech services or run supported transcription and speech generation in your own environment. Customer-hosted speech services can serve multiple agents and Rooms.
A customer-only speech setting keeps that path on the approved customer route. If the required service is unavailable, the voice operation stops instead of silently moving to shared processing.
Understand the processing routesKeep the useful part of the conversation.
Voice produces text that stays with the chat or Room under its retention rules. Audio is processed transiently in the voice path.
The speech route and the Agent’s reasoning model are separate choices. For work that must stay inside your environment, account for both, along with connected tools and other processing. External meeting connections are unavailable in strict-private configurations.