EVI (Empathic Voice Interface) by Hume AI is a speech-to-speech AI API that measures vocal tone, rhythm, and timbre to infer emotional state and generate contextually appropriate, expressive responses. It targets developers building voice assistants, customer service bots, and companion apps requiring emotional sensitivity. EVI integrates with major LLM backends including Claude, GPT, and Gemini, and supports voice cloning and custom voice design. Key features: - Real-time emotion detection from vocal prosody (tone, rhythm, timbre) - Speech-to-speech with sub-second latency and streaming audio output - Compatible with Claude, GPT-4, Gemini, Llama and other LLM backends - Voice cloning and natural-language voice design - Interruption handling and turn-taking for natural conversation flow
Free tier: ~5 min EVI usage/month. Paid plans from $3 to $500/month. EVI usage from $0.07/min (free overage) down to $0.05/min at scale.