Hume EVI, the Empathic Voice Interface, is a voice AI API for building agents that pick up on how a person sounds as well as what they say. It streams speech with about 300 milliseconds to first byte, lets users interrupt at any time, and works with an external language model of your choice. Voices come from a curated library, from cloning a sample, or from a natural-language description. Pricing starts with a free plan, followed by Starter at $3 per month, Creator at $14 (with the first month half off), Pro at $70, Scale at $200 and Business at $500.
The product fits developers, voice-product teams and companies that want conversational agents with emotional awareness: wellness and companion apps, language tutors, training programs and customer service. SDKs cover React, TypeScript, Python, Swift and C# (.NET), so most application teams can integrate it with a few lines of code. Enterprise buyers can rely on SOC 2 Type II and HIPAA compliance, and non-coders can test agents in an interactive playground before committing to development.
Case studies show how it gets used. Niantic Spatial built location-aware companions for Snap Spectacles AR glasses, and Coconote added voice chat so students can talk through their lecture notes and take quiz-style conversations. Other scenarios include a Spanish practice partner, a podcast-interview coach and a robo-butler style organizer. Developers inject dynamic variables such as user names or prices, add context mid-conversation for retrieval, call their own APIs for appointments or orders, and pull full transcripts with emotion data afterward.
