---
id: "b891"
language: "en"
library: "tools"
official_url: "https://huggingface.co/hexgrad/Kokoro-82M"
profile_updated_on: "2026-10-22"
---

# Kokoro

An 82M-parameter open-weight TTS model that reads text with language and voice selections and produces audio chunks through Python.

- Company / organization: hexgrad
- Category: Audio & music
- Official website: [https://huggingface.co/hexgrad/Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M)
- Profile updated: 2026-10-22


## Overview

Kokoro’s KPipeline accepts text, a language code and a preset voice. Misaki handles grapheme-to-phoneme processing, and generated audio chunks can be saved as WAV. Integration requires model and voice files plus language dependencies; this workflow selects provided voices rather than cloning arbitrary reference audio. Official weights use Apache 2.0. Language and voice inventories depend on the version, and the entry opens the author’s model card.

## Official sources and references

- [Official website](https://huggingface.co/hexgrad/Kokoro-82M)
- [Documentation](https://github.com/hexgrad/kokoro)

## Reading & subscriptions

- [Interactive profile](../../../products/kokoro--b891.html)
- [Read](../../../read/en/products/kokoro--b891.html)
- [JSON](../../../read/en/products/kokoro--b891.json)

Descriptions summarize public information; consult official websites for features, pricing and regional availability.