himaia

About

Voice infrastructure designed for character applications. A persona.yaml file establishes their identity. With a single API request, they can speak. This consistent persona operates across all scenarios—whether comfort, banter, or challenge—without requiring a fragile pipeline of LLM prompts, emotion tags, and TTS that fails with each update. The voice.persona project is Apache-2.0 licensed and available on GitHub. Treat it as a pinned dependency. Three distinct modes share a unified endpoint: input JSON, output WAV. Twenty voiced minutes are provided free each month, no credit card required.