Kokoro 82M

hexgrad

Ultra-lightweight TTS model. Under $1 per million characters. 54 pre-built voices across 8 languages. Apache 2.0 for commercial deployment. 8.9M+ monthly HuggingFace downloads.

Text-to-Speech Local Latest Kokoro Family v1.0

Documentation Back to Models

Type

Text-to-Speech

Source

Local

License

Apache 2.0

Capabilities

🌊

Streaming

Local Model Specs

Architecture

StyleTTS 2 + ISTFTNet vocoder

Runtime

Python / torch

Disk Size

0.3 GB

Details

Release Date: January 27, 2025
Knowledge Cutoff: -
Source: Local
License: Apache 2.0
Model ID: kokoro-82m

Last updated: March 13, 2026