Hume AI

Hume AI builds voice and emotion AI infrastructure with a distinguishing thesis: empathic interfaces that understand and respond to emotional tone, not just words. The product line covers EVI (Empathic Voice Interface — closed-source speech-to-speech with interruption handling and back-channeling), Octave (closed-source LLM TTS with voice design, modulation, cloning, and conversion), TADA (open-source LLM TTS that streams text and audio together to reduce hallucinations), and Expression Measurement (emotional intelligence analysis across 48+ emotions and 600+ voice descriptors). Curated speech datasets across 50+ languages and 10+ domains support training and benchmarking. Hume's positioning is the empathic-AI niche that ElevenLabs (voice cloning, no emotion-aware turn-taking), OpenAI Realtime API (voice, no expression measurement), Cartesia (low-latency TTS, no emotion layer), Vapi (voice agent infra), and Retell (voice agent platform) don't cover natively.