ltts

Quick CLI for local text-to-speech using Kokoro TTS.

Install

Recommended (fast, reproducible):

uv tool install ltts

Run without installing:

uvx ltts "hello world"

With pip:

pip install ltts

Usage

# Basic usage (outputs to output.mp3)
ltts "hello world"

# Specify output file
ltts "your text here" -o speech.mp3

# Different formats supported
ltts "test" -o output.ogg   # OGG
ltts "test" -o output.flac  # FLAC
ltts "test" -o output.wav   # WAV

# Use different voice
ltts "custom voice" -v am_adam    # Male American English
ltts "bonjour" -v ff_siwis        # French
ltts "こんにちは" -v jf_alpha      # Japanese

# Specify language code manually
ltts "こんにちは" -v jf_alpha -l j  # Japanese with explicit lang code

# Play audio through speakers instead of writing a file (ignores -o/--output)
ltts "Hello world" --say

# Stream audio chunks as they're generated (faster initial playback, useful for longer audio files)
# May have brief pauses between chunks on slower hardware
ltts "Hello world" --say --chunk

# Read text from stdin (pipe)
echo "Hello from pipe" | ltts --say
cat notes.txt | ltts -o notes.mp3

# See all available voices
ltts --help

Available Voices

Kokoro supports 50+ voices across multiple languages:

• American English: af_heart, af_alloy, af_bella, af_nova, af_sarah, am_adam, am_michael, and more
• British English: bf_alice, bf_emma, bf_isabella, bm_daniel, bm_george
• Japanese: jf_alpha, jm_kumo
• Chinese: zf_xiaobei, zm_yunxi
• Spanish: ef_dora, em_alex
• French: ff_siwis
• Hindi: hf_alpha, hm_omega
• Italian: if_sara, im_nicola
• Portuguese: pf_dora, pm_alex

Full voice list: https://huggingface.co/hexgrad/Kokoro-82M/blob/main/VOICES.md

Development

uv sync
uv run ltts "hello world"            # run the CLI using local code
uv run python -m ltts "hello world"  # alternative module form

# optional: editable install
uv pip install -e .

./scripts/release.sh                 # release a new version

Notes

• First run downloads the model (~330MB) to ~/.cache/huggingface/
• Japanese voices: First use automatically downloads the Japanese dictionary (~526MB one-time download)
• Supports MP3, OGG, FLAC, and WAV output formats
• Language code is auto-detected from voice prefix (or use -l to specify manually)
• --say notes:
  • Plays at 24 kHz on the default output device.
  • On Linux, ensure your user can access audio (PulseAudio/PipeWire running).
  • On macOS, you may see a permissions prompt for audio.