Kokoro TTS Container

A Docker container for running Kokoro Text-to-Speech engine v.1, providing high-quality speech synthesis with 54 voices and 9 languages options.

Container Features

• High-quality text-to-speech synthesis
• Multiple voice and languages options
• Voice blending capabilities
• Adjustable speech speed
• Support for .mp3 and .wav output files

Quick Start Using Docker Hub Image

You can directly pull and run the pre-built container from Docker Hub without building locally:

# Pull the latest image
docker pull usrbinbrain/kokoro-tts-container:latest

# Run a basic example
docker run --rm -v $(pwd):/app/shared usrbinbrain/kokoro-tts-container \
    "Hello world!" \
    output.mp3 \
    --voice "af_sarah" \
    --speed 1.0 \
    --lang "en-us"

This way you can use Kokoro-TTS instantly without worrying about setup or build steps.

Local Setup && Build

Building your kokoro-tts Docker image:

# Install requirements for setup
pip3 install -r requirements.txt

# Run setup to donwload model and gerenate voices bin file
python3 setup.py

# Build your kokoro-tts image
docker build -t kokoro-tts-container .

Usage

Basic Usage Examples

Run the container with a single voice.

The command below generates an output.mp3 file, where af_sarah voice says "Hello my friend!" in English (US) with speed 1.2

docker run --rm -v $(pwd):/app/shared kokoro-tts-container \
    "Hello my friend!" \
    output.mp3 \
    --voice "af_sarah" \
    --speed 1.2 \
    --lang "en-us"

Voice Blending

Kokoro-TTS supports voice blending, allowing you to mix multiple voices with different weights.

The command below generates an output.wav file with combined voices, where af_sarah contributes 40% and am_adam contributes 60% to the final voice saying "Hasta la vista!" in Spanish with speed 0.8

docker run --rm -v $(pwd):/app/shared kokoro-tts-container \
    "Hasta la vista!" \
    output.wav \
    --voice "af_sarah:40,am_adam:60" \
    --speed 0.8 \
    --lang "es"

Container Parameters

Parameter	Description	Default
input_text	The text to synthesize	Required
output_file	Output audio filename (.wav or .mp3)	Required
--voice	Voice ID or blend (format: voice1:weight,voice2:weight)	af_sarah
--speed	Speech rate multiplier, allows 0.5 to 2.0	1.0
--lang	Language code	en-us

Supported Languages and Codes

• en-us: English (US)
• en-gb: English (British)
• fr-fr: French
• ja: Japanese
• hi: Hindi
• cmn: Mandarin Chinese
• es: Spanish
• pt-br: Brazilian Portuguese
• it: Italian

Available Voices

The container includes multiple voices for different languages, for a complete list of voices or another help, run:

docker run --rm kokoro-tts-container --help

Thanks

Built with ❤️ on top of Kokoro ONNX - A special thanks to thewh1teagle and hexgrad for providing this amazing fast TTS engine that made this container project possible.