Bottom line: Kokoro-82M is completely free text-to-speech with no paid version at all. It is an open-weight model released under the Apache 2.0 license: download it, run it on your own computer — it is small enough to run on a CPU — and generate as much narration as you like, offline and private. The honest trade-off is that it cannot clone voices; you choose from its built-in voices.
What is Kokoro?
Kokoro-82M is an 82-million-parameter text-to-speech model created by hexgrad and published on Hugging Face. Version 1.0, released January 27, 2025, ships 54 voices across 8 languages in weights of roughly 350 MB. Browser demos and community tools wrap the same model, but the core offer is simple: local, free, unlimited speech synthesis.
What is actually free with Kokoro?
- Completely free and open source (Apache 2.0): download and run locally with no usage caps, no account, and no paid version.
- 54 voices across 8 languages in the v1.0 release.
- Small enough to run on CPU — no GPU required, and it works offline after download.
- Commercial use is allowed under the Apache 2.0 license.
- Private by design: your text never leaves your computer when you run it locally.
What is the catch?
There is no hosted studio, no support desk, and no voice cloning. You (or a community web demo) provide the interface and the hardware. Kokoro speaks with its preset voices only, and dramatic or highly emotional delivery is not its strength — it is built for clean, natural narration. Hosted demos built on Kokoro may set their own limits; the model itself has none.
Pros and cons
Pros
- Truly $0 forever: no credits, no per-character fees, no subscription tier.
- Runs on ordinary hardware, including CPU-only machines, and works offline.
- Apache 2.0 license permits commercial use.
- Privacy: scripts and documents stay on your device.
Cons
- No voice cloning — preset voices only.
- Setup is on you; there is no polished paid-style studio app from the maker.
- Less expressive range than top commercial voice services.
Who is Kokoro best for?
Kokoro is best for creators narrating videos, audiobooks, and lessons on a budget, developers embedding speech into apps without API bills, and privacy-conscious users who want text read aloud without sending it to a cloud service.
FAQ
Is Kokoro really free for commercial use?
Yes. The model weights are released under the Apache 2.0 license, which permits commercial use, as long as you follow the license terms.
Do I need a GPU for Kokoro?
No. At 82 million parameters it runs comfortably on CPU, which is a big part of its appeal.
Can Kokoro clone my voice?
No. Kokoro offers its built-in voice set only. If cloning matters, look at an open model built for cloning or a paid voice service.
Does Kokoro work offline?
Yes, when run locally — after the one-time model download, generation happens entirely on your machine.
Research-based review: license, voice, and language details checked against the official Hugging Face model card in October 2026. No hands-on testing is claimed.