Hume AI EVI 2

Hume AI EVI 2 is the second-generation Empathic Voice Interface, announced in September 2024, delivering a 40 percent latency reduction versus EVI 1 and a 30 percent price reduction to $0.072 per minute. It is a native voice-to-voice model that integrates speech sentiment analysis directly into the generation loop, allowing the agent to modulate tone, pacing, and prosody in real time based on the detected emotional state of the user. This design made it the first commercially available speech API with true end-to-end empathic response at launch. EVI 2 ships with configurable voice parameters that allow developers to shape expressiveness, warmth, and energy without requiring full voice cloning. Full voice cloning from short audio samples is available on higher-tier plans. The model supports a bring-your-own-LLM interface, letting teams plug in GPT-4o, Claude, or Llama while retaining EVI's emotional intelligence wrapper. Function calling and barge-in detection are both supported natively over a WebSocket connection. EVI 2 is the right fit for companion app developers building wellness assistants, mental health tools, or social AI products where vocal empathy defines the user experience. At $0.072 per minute, it costs more than commodity TTS cascade APIs but delivers a qualitatively different interaction layer that cannot be replicated by generic LLM audio models. Its successor EVI 3 offers further latency and accuracy improvements; EVI 2 remains available as a stable lower-cost option for teams already integrated against it.