I built a local TTS add-on using an 82M parameter neural model – Runs on potatos

4 months ago 12

🧠 Local Neural Text-to-Speech for Firefox — fast, private, offline.

Tested on a Xeon E3-1265L v3 (2013) — Ran multiple TTS jobs in parallel with barely perceptible lag.
If it works on this, it'll fly on your machine.

Kokoro TTS is a Firefox extension that lets you convert selected or pasted text into natural-sounding speech — without needing an internet connection.
It uses a lightweight Flask server and the Kokoro model running locally on your system.

✅ No accounts or logins
✅ No cloud APIs or telemetry
✅ No GPU required but helps a lot, if no usable GPU falls to using the CPU.

🎙️ Neural TTS with multiple voice options
🔒 Offline-first & privacy-respecting
🧊 Lightweight: Small 82M parameters
🥔 Works on low-end CPUs
🌍 Linux, macOS, and Windows support

1. Download from Releases

Head to the Releases Page and grab:

latest kokoro-tts-addon.xpi
server.py

2. Install the Add-on in Firefox

Go to about:addons
Click the gear icon → Install Add-on From File...
Select the .xpi you downloaded

3. Start the Local Server

nohup python3 /path/to/server.py &

Create a .bat file like this:

cd C:\path\to\server start python server.py

Drop a shortcut to it in the Startup folder (Win + R → shell:startup).

To install espeak-ng on Windows:

Go to espeak-ng releases
Click on Latest release
Download the appropriate *.msi file (e.g. espeak-ng-20191129-b702b03-x64.msi)
Run the downloaded installer

For advanced configuration and usage on Windows, see the official espeak-ng Windows guide

Visit http://localhost:8000/health
You should see a simple “healthy” JSON response
Use the extension: paste text, pick a voice, click “Generate Speech” 🎉

First-time run will download the model
Make sure Python 3.8+ is installed and in PATH
All processing is local — nothing leaves your machine

You’ll need Python 3.8+ and pip installed. Most systems already have them.
To install all required Python packages (including some optional extras for extended model usage), run:

pip install --upgrade pip setuptools cat requirements.txt | xargs -n 1 pip3 install pip3 install -U flask-cors

Licensed under the Apache License 2.0