# Hermoso: Licenses and Credits for the Free AI Voice Generator

> Machine-readable twin of https://hermoso.ai/tools/free-ai-voice-generator/licenses/. Generated from the shipped page. Do not edit by hand.
> Full agent guide: https://hermoso.ai/llms.txt · MCP: https://app.hermoso.ai/mcp

# Licenses, credits and source

The free voice generator is built from open models and code. Here is each part, its license and where to get the source.

The free AI voice generator is built from open-source models and code. This page lists each part, its license, and where to get the source. All files are served unmodified except where noted.

Part License Source

Kokoro-82M v1.0 voice model (ONNX build by the Transformers.js community, 8-bit quantized) and its voice files Apache License 2.0 [huggingface.co/hexgrad/Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M), [huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX](https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX)

kokoro-js 1.2.1 Apache License 2.0 [github.com/hexgrad/kokoro](https://github.com/hexgrad/kokoro)

Transformers.js 3.5.1 Apache License 2.0 [github.com/huggingface/transformers.js](https://github.com/huggingface/transformers.js)

ONNX Runtime Web 1.22 (the WebAssembly engine) MIT License [github.com/microsoft/onnxruntime](https://github.com/microsoft/onnxruntime)

phonemizer.js 1.2.1 (turns text into sounds) Apache License 2.0 [github.com/xenova/phonemizer.js](https://github.com/xenova/phonemizer.js)

**eSpeak NG**, compiled to WebAssembly and embedded in phonemizer.js **GNU GPL version 3 or later** [github.com/espeak-ng/espeak-ng](https://github.com/espeak-ng/espeak-ng)

## eSpeak NG and the GPL

phonemizer.js labels itself Apache-2.0, but the speech engine it embeds, eSpeak NG, is licensed under the GNU General Public License, version 3 or later, and the downloaded script contains it in compiled form. We therefore offer the whole in-browser voice program under the GPL-3.0 terms. You may copy, study, change and redistribute it under that license.

- The complete license text: [GPL-3.0](https://hermoso.ai/assets/voice/LICENSE-GPL-3.0.txt).

- eSpeak NG source: [github.com/espeak-ng/espeak-ng](https://github.com/espeak-ng/espeak-ng). The WebAssembly build comes from [phonemizer.js](https://github.com/xenova/phonemizer.js) (see its `src/espeakng.worker.js` and build scripts).

- The script served at `/assets/voice/kokoro-web.22b79dc3.js` is the unmodified kokoro-js 1.2.1 browser bundle with only the download addresses changed to assets.hermoso.ai. The page script that drives it, [voice-tool.070fec0f.js](https://hermoso.ai/assets/voice/voice-tool.070fec0f.js), and the worker that runs the model, [voice-worker.0fd56bf4.js](https://hermoso.ai/assets/voice/voice-worker.0fd56bf4.js), are plain readable JavaScript, released under the MIT license.

- For a copy of any source on physical media, or if a link here is broken, write to hello@hermoso.ai and we will send it.

## Other license texts

- [Apache License 2.0](https://hermoso.ai/assets/voice/LICENSE-Apache-2.0.txt) (Kokoro, kokoro-js, Transformers.js, phonemizer.js)

- [ONNX Runtime, MIT License, Copyright (c) Microsoft Corporation](https://hermoso.ai/assets/voice/LICENSE-onnxruntime-MIT.txt)

## Credits for the voice training data

The Kokoro model was trained on permissive and non-copyrighted audio. Its authors list two Creative Commons Attribution datasets that were part of the training set for version 1.0: the Koniwa `tnc` corpus (CC BY 3.0) and the SIWIS French speech synthesis database (CC BY 4.0). Kokoro was architected on StyleTTS 2 by Li et al. and trained by the Kokoro author; see the [model card](https://huggingface.co/hexgrad/Kokoro-82M) for details.

## Changes we made

The download addresses inside the kokoro-js bundle were changed from Hugging Face and jsDelivr to assets.hermoso.ai so that everything is served from our own storage. The ONNX Runtime WebAssembly file and its script glue are the files published with Transformers.js 3.5.1, unchanged. The model and voice files are the 8-bit ONNX build, unchanged.

## Frequently asked questions

### Is the free AI voice generator open source?

It is built from open-source parts: the Kokoro model and Transformers.js under Apache-2.0, ONNX Runtime under MIT and eSpeak NG under the GPL. The page script is readable JavaScript under the MIT license.

Back to the free AI voice generator, or start free at Hermoso.

[Start free →](https://app.hermoso.ai/?src=tool-free-ai-voice-generator-licenses)

[See pricing](https://hermoso.ai/pricing)