Hume AI Octave

Octave is Hume AI's Omni-Capable Text and Voice Engine (OCTAVE): a voice-based large language model that generates emotionally nuanced, context-aware, and human-like speech.

Octave is not a standard text-to-speech (TTS) model; it is a proprietary voice-based LLM designed for expressive communication. The engine interprets text context to predict emotion, cadence, and sarcasm, then delivers speech with unprecedented realism. It generates complete personalities—not just voices—from simple descriptive prompts, such as a 'gentle therapist' or a 'dramatic medieval knight.' Octave can also extract and clone a speaker's full vocal identity and personality from a recording as brief as five seconds. In blind comparison tests, 180 human raters favored Octave's outputs over a competitor's for audio quality (71.6%) and naturalness (51.7%), confirming its position as a frontier voice technology.

https://hume.ai