Hume

Developing artificial intelligence systems that prioritize human emotional well-being. Real human ratings, in a single API call. The human evaluation layer for voice, speech, and conversational AI. Hume builds voice AI with built-in emotional intelligence. Its core offerings like Octave (text-to-speech) and EVI (speech-to-speech Empathic Voice Interface) understand context and tone to produce realistic, expressive voices that react to human emotions. Unlike standard TTS systems, Hume uses LLM-based models (especially Octave) that grasp meaning and emotion in text or speech. This leads to more natural cadence, empathy, and nuance, particularly in conversational or expressive scenarios. EVI stands for Empathic Voice Interface. It's a speech-to-speech foundation model that listens to tone, rhythm, and expression in real time, then responds with matching emotional intelligence, low latency under 300ms, and human-like interruptions or empathy. Octave 2 and EVI support 11+ languages including English, Spanish, French, German, Japanese, Korean, Portuguese, Italian, Russian, Hindi, and Arabic, with the same emotional expressiveness across them. You can design voices from prompts, clone existing ones, or pick from their Voice Library, then use them in TTS or conversational EVI setups for things like characters or branded agents. Especially for apps needing empathy. Developers integrate EVI via API to create voice companions, customer support bots, game NPCs, or phone agents that sound caring and respond naturally to frustration, excitement, sadness, and more.