Podium Voices: multi-agent AI hosts for live audio rooms (turn coordination) (github.com)

Podium has announced the launch of Podium Voices, a minimum viable AI co-host for its Outpost audio rooms, integrating multi-agent AI capabilities designed for live audio interactions. This AI agent employs a modular system that facilitates live speech transcription (ASR), response generation using large language models (LLM), and speech synthesis via text-to-speech (TTS) technologies. Users can select from different backends, such as the standard ASR-LLM-TTS pipeline or the new PersonaPlex speech-to-speech system, allowing each agent in a multi-agent setup to utilize unique configurations and capabilities. This development is significant for the AI/ML community as it introduces a more flexible and dynamic framework for creating interactive audio experiences, potentially revolutionizing virtual events and podcasts. The ability to configure agents with various toolsets and personalizations enhances the user experience, enabling agents to respond in varied tones or styles, such as more natural or influencer-like voices. By allowing multiple AI agents in a single room to communicate seamlessly while an overarching Turn Coordinator manages the interactions, Podium Voices paves the way for richer, more engaging discussions in augmented audio environments.