Higgs AudioUnclaimed

Boson AI builds voice agents and the speech models behind them. Its main model, Higgs Realtime, is a speech-to-speech model that listens, reasons, calls tools and answers in its own voice, with nothing lost in transcription. The page says it is drop-in compatible with the OpenAI Realtime API and reports about 0.7 seconds from speech in to speech out.

The company also offers Higgs Audio for expressive voices that can be cloned or designed, Higgs Avatar for a real-time talking face driven by audio, and Higgs TTS 3 for speech generation in more than 100 languages. Boson Workspace is a single console for every service.

The page shows businesses using it for sales calls, insurance sales specialist training, customer support and AI receptionists. Agents can be built with state machines, knowledge bases, tool calling and granular control, and the company says they can be tailored to a domain and trained on a business's own data.

Pricing is usage-based. Higgs Realtime is listed at $0.0023 per minute of audio input and $0.014 per minute of audio output, Higgs TTS 3 at $0.015 per 1K characters, and Higgs Avatar from $0.07 per minute. New accounts get $10 in free trial credit, and an API key can be requested without commitment.