From the archive
ElevenLabs introduces configurable voice agents
Conversational AI combines speech recognition, voice generation, and a language model with turn-taking and tool connections.
ElevenLabs introduced its Conversational AI platform in a December 3, 2024 announcement. The service combined speech recognition, speech generation, a language model, and conversation handling for interactive voice agents.
More than generating a voice
The platform handled turn-taking and interruptions as well as the spoken reply. Agent configuration included a knowledge base, instructions, tools, and a choice of voice. Examples included scheduling, customer support, tutoring, and characters in games.
Connections to an application
ElevenLabs offered software development kits and a WebSocket interface, with the option to connect an external language model or server. The announcement also described a Twilio integration for phone calls and tool calls on the server or client.
The release expanded ElevenLabs from producing audio into managing a conversation that could retrieve information or trigger an action.