Hume AI

Hume AI is an open-source data and evaluation platform for developers building emotionally intelligent voice AI. It exposes an API layer covering expression measurement, conversation simulation, and human feedback collection — giving engineering teams the instrumentation needed to evaluate and iterate on voice AI systems with emotional fidelity.

Key technical capabilities include:

• Expression Measurement: 48+ emotions, 50+ languages, and 600+ output metrics for granular signal extraction from voice and multimodal inputs
• Empathic Voice Interface (EVI): Real-time expression measurement integrated into voice AI interactions
• Kairos Simulation: Agent-to-agent and human-to-agent conversation simulation for scalable, automated evaluation pipelines
• Human Feedback at Scale: Pre-screened rater panels for collecting ground-truth human judgment data
• Voice AI Leaderboards: Benchmarks grounded in human judgment for comparing model performance
• External LLM Integrations: Connect your own LLM backends into the evaluation and voice pipeline

The project is open source with 1,016 GitHub stars and offers a freemium pricing model. Usage-based overages apply for TTS characters and EVI minutes, with plans ranging from a free tier up to enterprise custom pricing.