How to run Kokoro-82M locally

hexgrad · Text → speech · 82M params · Apache-2.0 (commercial OK)

A remarkably tiny 82M TTS model with quality rivaling far larger systems, Apache-licensed and runnable nearly anywhere.

What it costs to run — $0 markup

On your own machine — $0. Runs free if you have about 1.0 GB of VRAM (Tiny; runs real-time on CPU or <1GB VRAM), via Kokoro / ONNX.
Rent a GPU — from $0.26/hr. Fits RTX 3090 24GB at the direct vendor price ($0 markup) — pay only for the minutes you generate.
Download the weights — free. Open weights at hexgrad/Kokoro-82M.

Note: generative-media models are billed per image / per second / per minute on hosted services — not per token. Running locally or on your own rented GPU is usually far cheaper and keeps your data on your machine.

Key facts

Does	Text → speech
VRAM to run	~1.0 GB (Tiny; runs real-time on CPU or <1GB VRAM)
Download	~0.35 GB
Parameters	82M
License	Apache-2.0 (commercial use OK)
Run with	Kokoro / ONNX

Get Kokoro-82M on Hugging Face →

More voice & audio models

Whisper large-v3 — Speech → text, ~10 GB VRAM
Whisper large-v3-turbo — Speech → text, ~6.0 GB VRAM
XTTS-v2 — Voice cloning, ~6.0 GB VRAM
MusicGen Large — Text → music, ~16 GB VRAM
Chatterbox — Text → speech, ~8.0 GB VRAM
Parakeet TDT 0.6B v2 — Speech → text, ~4.0 GB VRAM

Browse: all media models · chat / LLM models