← toolbox

Murmur-Kokoro

v0.1.0

A private voice pack for Murmur that adds natural-sounding Kokoro voices running inside Chrome. Kept for private use only and not offered for download; this page explains why and where the pieces come from.

projectmini tool

At a glance

Status
work in progressActively changing; expect things to move.
Being worked on?
Being worked on
Version
v0.1.0
Shelf
Ideas and placeholders
Added
Last updated
Known bugs
None noted
Runs
An idea, not built yet
Saves your data
Nothing yet

What it does

What's next

The write-up

Murmur-Kokoro

Murmur-Kokoro is a private voice pack for Murmur, the Chrome extension that reads highlighted text aloud. It adds the Kokoro voices: natural-sounding English voices that run on the computer itself, inside Chrome, with no internet connection needed once set up. It is built and used privately, and it is not offered for download from this site.

Why it is not offered for download

The speech engine that makes Kokoro work includes eSpeak NG, a free program under the GNU General Public License (GPL-3.0). That licence has conditions for anyone who hands out copies of software containing it. To keep things simple, this pack is kept for private use and is not distributed. Nothing here is a way to download it.

Where the pieces come from

If you would like Kokoro voices for yourself, these are the original sources, all free to get from their own authors:

Murmur itself is free to download from its own page and works with the voices already on your computer. Another way to hear text read aloud on a phone, with no install, is MurmurMini.

Saving and privacy

This page only describes the tool and saves nothing. The pack itself, on the computer where it is installed, keeps its settings in Chrome's extension storage and the model files in the browser's cache for the extension. It has no network access, no access to websites, tabs or files, and only answers Murmur.

Current state

Work in progress and private. Built and tested end to end with a stand-in engine; the real voice is yet to be measured on real hardware.

Changelog