[ STACK LIBRARY ]

Hexgrad (Kokoro)

Hexgrad (Kokoro)

Visit website

Text-to-Speech

ABOUT

Open-weight, lightweight text-to-speech model with 82 million parameters and Apache-licensed weights

Kokoro uses a StyleTTS 2 and ISTFTNet architecture and is decoder-only, with no diffusion. Version 1.0 supports 8 languages and 54 voices. It was trained on a few hundred hours of permissive, public domain or synthetic audio for about $1,000 of A100 compute. Its authors say it matches the quality of larger models while being faster and cheaper to run. It installs as the open-source kokoro Python package, runs anywhere from personal projects to production, and is also served by third-party APIs, which the model card says charge under $1 per million characters. The Hugging Face page shows over 10 million downloads last month.

STACK COVERAGE

Text-to-Speech

DETAILS

FOUNDED

2024

Is this your company? To claim this listing or update its details, contact us at sunil@uniocommunity.com

UNIOVOICE AI STACK LIBRARY
Hexgrad (Kokoro) logo
Trusted by Unio

Hexgrad (Kokoro)

Text-to-Speech

Open-weight, lightweight text-to-speech model with 82 million parameters and Apache-licensed weights

Layer
Text-to-Speech
Market
Global
Founded
2024
Listed on the Unio Voice AI Stack Library

More in this layer

var(--variable-H7lUBNgYB)

Pradhi

Text-to-Speech

Advanced voice intelligence for BFSI that reads intent from tone, not just transcripts.

var(--variable-H7lUBNgYB)

SpeechRouter

Text-to-Speech

SpeechRouter is a speech AI infrastructure platform that gives developers access to multiple speech models through one API. It routes transcription requests between models by language or task, so apps don’t depend on a single provider.

var(--variable-H7lUBNgYB)

AI-coustics

Text-to-Speech

Real-time speech enhancement that cleans audio so voice AI transcribes and responds reliably.