Hume AI

Listed at https://hume.ai

Overall Rank #28 ⭐ Consider
Proxy required (US export controls) | 🌍 International

💰 Token Pricing

TypePriceNote
Input EVI 3: ~$0.096/min (pay-per-second); OCTAVE TTS: $0.048/1K chars; Expression Measurement: $0.0008/sec per million tokens
Output Same as input (most endpoints symmetric) per million tokens
💡 Free Credits: Free credits on signup (amount per account), enough to trial EVI 3

🤖 Supported Models (5)

EVI 3 (Empathic Voice Interface 3)OCTAVE TTS (voice design + cloning)Expression MeasurementSpeech-to-Speech (EVI 2/3)Voice Design (text -> custom voice)

✨ Pros

  • EVI 3 (Empathic Voice Interface): real-time voice conversations that detect and respond to user emotions
  • OCTAVE TTS: industry-first TTS model that generates custom voices from natural-language descriptions (e.g. male, deep, calm)
  • Expression Measurement: quantifies 48 emotional dimensions from voice, video, or image
  • Voice Design: generates entirely new voices from text prompts, no reference audio required
  • MCP integration (Jan 2026): Hume's official MCP server lets Claude and Cursor call EVI / OCTAVE natively

⚠️ Cons

  • ×Pricing is higher than text-only TTS (e.g. Cartesia) -- premium for empathic capabilities
  • ×China access requires proxy; latency is 200-400ms from Shanghai
  • ×OCTAVE TTS currently supports English and limited multilingual coverage
  • ×EVI real-time conversations require WebSocket protocol, integration cost for legacy clients

🎯 Best For

Empathic voice agents (customer service, companionship, therapy, coaching); multimodal products that need to read user emotion; TTS where you describe the voice in natural language rather than clone a reference recording