Emotion-Adjustable AI Voices

Hume has introduced Octave, a new text-to-speech AI model designed to generate lifelike and emotionally nuanced speech. Unlike conventional text-to-speech systems that merely read words, Octave is powered by a large language model (LLM) trained on both text and speech, enabling it to understand words in context and adjust tone, rhythm, and cadence accordingly. This allows Octave to produce more natural and expressive speech, making it suitable for various applications such as audiobooks, podcasts, video game characters, and video voiceovers. Users can also customize the generated voice by providing text instructions to adjust emotions and styles, such as making the voice sound happier, sadder, or more sarcastic.

Octave's capabilities extend beyond basic voice generation, as it can interpret character traits and styles from a script alone, adjusting vocal inflections to match implied emotions. For example, a sarcastic remark will be spoken sarcastically, and a panicked sentence will sound urgent. The model also supports multiple languages, including English and Spanish, with plans to expand its language capabilities in the future. Hume offers access to Octave through its website and API, allowing developers to integrate the model into their projects and create realistic, character-specific voices for various content types.

Image Credit: Shutterstock