VoxKit

A Python library for managing Piper TTS voice models. Browse, install, and use offline AI voices with a clean API and an interactive terminal UI.

Features

• Voice catalog — list, search, and filter Piper voices by language or locale
• Install/uninstall — download voice models to any directory you choose
• TTS synthesis — synthesize text to WAV files with volume control
• Interactive TUI — terminal-based voice browser with install, uninstall, and test
• Configurable — custom data directories, catalog URLs, cache TTL, and TUI keybindings
• Offline — all synthesis runs locally, no internet needed after voice download

Requirements

• Python 3.10+
• Linux (uses aplay for WAV playback, mpg123 for MP3)
• Piper TTS (installed automatically as dependency)

Installation

pip install voxkit

Quick Start

from voxkit import VoiceManager

# Initialize with your app's storage directory
vm = VoiceManager(data_dir="~/.myapp/voices")

# List available English voices
voices = vm.list_voices("en")
for v in voices:
    print(f"{v['key']}  {v['quality']}  {v['size_mb']:.0f} MB  {'[installed]' if v['installed'] else ''}")

# Install a voice
vm.install("en_US-lessac-medium")

# Synthesize and play text
vm.speak("en_US-lessac-medium", "Hello, world!")

API Reference

VoiceManager

VoiceManager(
    data_dir,             # Directory for models and cache
    catalog_url=None,     # Override Piper catalog URL
    base_url=None,        # Override download base URL
    cache_ttl=86400,      # Catalog cache TTL in seconds (default: 24h)
)

Catalog

vm.list_languages()                # -> ['en', 'pl', 'de', ...]
vm.get_language_name("en")         # -> 'English'
vm.list_voices("en")               # all English variants
vm.list_voices("en_US")            # US English only
vm.get_voice("en_US-lessac-medium") # single voice info or None

The lang parameter accepts both 2-letter codes ("en") to match all regional variants and full locale strings ("en_US") for exact matching.

Install & Manage

vm.install("en_US-lessac-medium", progress_cb=None)
vm.uninstall("en_US-lessac-medium")  # -> True if removed
vm.is_installed("en_US-lessac-medium") # -> bool
vm.get_path("en_US-lessac-medium")   # -> Path or None

The progress_cb receives (filename, block_num, block_size, total_size) during download.

Resolve

# Find best installed voice for a language
path = vm.resolve(lang="en")

# Use a specific voice (auto-downloads if missing)
path = vm.resolve(voice="en_US-lessac-medium")

# Prefer exact locale, fall back to any match
path = vm.resolve(lang="en_US")

TTS Synthesis

# Synthesize to WAV file
wav_path = vm.synthesize("en_US-lessac-medium", "Hello!")
wav_path = vm.synthesize("en_US-lessac-medium", "Hello!", output_path="out.wav")

# Multiple texts with pauses
wav_path = vm.synthesize_multi("en_US-lessac-medium", ["Hello.", "World."], pause_ms=700)

# Synthesize and play immediately
vm.speak("en_US-lessac-medium", "Hello, world!")

# Volume control
vm.set_volume(75)

Voice Loading

# Load a PiperVoice instance for direct use
voice = vm.load_voice("en_US-lessac-medium")

Language Detection

# Detect language from system locale
lang = VoiceManager.detect_language()  # -> 'en_US' or 'pl_PL' etc.

Interactive TUI

# Basic usage — browse English voices
vm.browse(lang="en")

# With configuration
from voxkit import BrowserConfig

config = BrowserConfig(
    lang="pl",
    default_voice="pl_PL-darkman-medium",
    show_size=True,
    test_fn=lambda key: vm.speak(key, "Dzień dobry"),
    on_install=lambda v: print(f"Installed {v['key']}"),
)
vm.browse(config=config)

The TUI supports:

• Arrow keys — navigate the voice list
• Enter — test the selected voice (auto-installs if needed)
• i — install the selected voice
• u — uninstall the selected voice
• q / Esc / Ctrl-C — quit

Keybindings are configurable via BrowserConfig.keybindings.

Built-in Progress Bar

VoxKit includes a ready-made terminal progress bar for downloads:

from voxkit.browser import progress_bar

vm.install("en_US-lessac-medium", progress_cb=progress_bar)

Data Types

from voxkit import VoiceInfo, BrowserConfig, ProgressCallback

VoiceInfo

Returned by list_voices() and get_voice() as a dict with keys: key, name, language, quality, region, speakers, size_mb, installed.

BrowserConfig

Default keybindings:

{
    "install": "i",
    "uninstall": "u",
    "test": "\r",
    "quit": ("q", "\x03", "\x1b"),
}

Storage Layout

<data_dir>/
├── models/          # Downloaded .onnx voice models
│   ├── <voice>.onnx
│   └── <voice>.onnx.json
└── cache/
    └── voices.json  # Cached Piper catalog (auto-refreshed)

License

Copyright (c) 2026 Jakub T. Jankiewicz

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

See LICENSE for the full text.