Hume Launches EVI 3 Empathic Voice AI for Natural Conversations

Hume, a New York-based AI startup, has unveiled its latest Empathic Voice Interface model, EVI 3, designed to revolutionize conversational AI with naturalness, expressiveness, and emotional intelligence. Pronounced “Evee Three,” this voice-to-voice AI allows users to create custom voices simply by speaking to the model, making interactions feel more human and empathetic.

EVI 3 targets a broad range of applications including customer support systems, health coaching, immersive storytelling, and virtual companionship. Unlike traditional voice assistants that often feel scripted or mechanical, EVI 3 adapts dynamically to speech patterns such as pitch, pauses, and prosody, creating conversations that are engaging and emotionally resonant.

One standout feature is the ability to specify detailed personality traits, vocal qualities, and emotional tones, allowing the AI to embody anything from a warm, empathetic guide to a quirky narrator. This flexibility opens doors for developers and creators to craft unique voice experiences tailored to their audiences.

While EVI 3 currently does not support voice cloning—a feature that allows rapid replication of specific voices—Hume plans to introduce this capability soon in its Octave text-to-speech engine, emphasizing ethical safeguards before broad release.

Internal benchmarks from Hume’s testing with over 1,700 users show EVI 3 outperforms OpenAI’s GPT-4o voice model and Google’s Gemini in naturalness, expressiveness, empathy, response speed, and audio quality. It also supports multilingual interactions and boasts low latency, making it suitable for real-time applications.

Key Features and Capabilities

• Expressive text-to-speech with prosody and emotional modulation
• Interruptibility for dynamic conversational flow
• Real-time voice customization during conversations
• API-ready architecture for seamless integration into apps and services (coming soon)

Pricing and Access for Developers

Hume offers flexible, usage-based pricing for its APIs. While EVI 3’s specific API pricing is yet to be announced, previous versions show competitive rates with discounts for enterprise customers. The Octave TTS engine provides free and tiered plans, making it accessible for creators of all sizes.

• Free tier: 10,000 characters, unlimited custom voices
• Starter to Enterprise plans scaling from $3/month to custom pricing with advanced support

The Vision Behind Hume’s Empathic AI

Founded in 2021 by former Google DeepMind researcher Alan Cowen, Hume focuses on bridging human emotional nuance with AI interaction. Their models are trained on extensive datasets capturing speech, vocal bursts, and facial expressions to enhance emotional intelligence in AI. This mission aims to make AI interfaces more responsive, natural, and useful across industries.

With EVI 3 now available for public demos and API access launching soon, developers and businesses have a powerful tool to create voice experiences that truly resonate emotionally and engage users like never before.