VoiceConductor 0.1.0

Importable text-to-speech package with Windows audio routing support.

VoiceConductor

A Python package for generating and routing synthesized voice lines. Supports output to speakers or virtual microphones.

Entry point: TTSManager

Features

• Provider fallback across ElevenLabs, Kokoro, Azure Speech, Windows Speech, and the built-in demo provider.
• SQLite phrase caching so repeated lines do not need to be synthesized again.
• Named playback routes for speakers and virtual mic devices.
• Background playback tasks for non-blocking speech.
• Playback lifecycle hooks for push-to-talk workflows.
• JSON/JSONC configuration for provider credentials, default voices, route settings, and cache paths.

Requirements

• Python 3.11 or newer.
• Audio playback support via sounddevice.
• Optional provider dependencies and credentials depending on the backend you want to use.

Installation

Install the package:

pip install VoiceConductor

Install with Kokoro support:

pip install "VoiceConductor[kokoro]"

For local development from a checkout:

pip install -e ".[kokoro]"

Quick Start

from voice_conductor import TTSManager

tts = TTSManager()
tts.speak("This is a test.", routes="speakers")

Route to a virtual mic:

from voice_conductor import TTSManager

tts = TTSManager()
tts.speak("Now, to the virtual microphone.", routes="mic")

Route to both speakers and mic:

tts.speak("Routed to both output devices.", routes=["speakers", "mic"])

Synthesize once, then route the resulting audio:

audio = tts.synthesize_voice("This audio sample is stored in audio.")
result = tts.route(audio, routes=["speakers", "mic"])
print(result.routes)

Configuration

By default, voice_conductor looks for one of these files in the current working directory:

• voice_conductor.config.jsonc
• voice_conductor.config.json

If neither file exists, defaults are used. To create a config file you can edit, save the current settings:

from voice_conductor import load_settings

settings = load_settings()
settings.save_settings("voice_conductor.config.jsonc")

Provider selection follows voice_conductor.provider_chain. When speak() or synthesize_voice() does not specify a provider, the manager uses the first available provider in that chain.

Providers

Built-in providers:

Provider	Use case	Availability
elevenlabs	Hosted high-quality voices.	Requires API key.
kokoro	Local Kokoro synthesis.	Requires the kokoro extra and model access.
azure	Azure neural voices.	Requires Speech key and region.
windows	Installed Windows System.Speech voices.	Requires Windows speech support.
demo	Offline test voice.	No external service.

List available providers:

from voice_conductor import TTSManager

tts = TTSManager()
print(tts.list_providers())

List voices for a provider:

for voice in tts.list_voices("windows"):
    print(voice.id, voice.name)

Audio Routes

Routes are named outputs. The default route names are:

• speakers
• mic

You can pass one route name or a list of route names:

tts.speak("Hello, world!", routes="speakers")
tts.speak("Hello, world!", routes=["speakers", "mic"])

The mic route resolves to an output device because virtual microphone tools such as VB-CABLE and VoiceMeeter expose playback endpoints that chat apps receive as microphone input. See docs/mic-setup.md for a virtual microphone setup guide.

Cache

Synthesized phrases are cached in SQLite. Cache entries are keyed by:

• text
• provider
• normalized voice key
• provider settings that affect audio output

Useful cache methods:

tts.invalidate_synthesis_cache(text="Hello, world!")
tts.invalidate_synthesis_cache(provider="elevenlabs")
tts.clear_synthesis_cache()

Pass refresh_cache=True to regenerate a phrase and replace the cached entry:

tts.speak("New take.", refresh_cache=True)

Background Playback

Use background=True when the caller should continue immediately:

task = tts.speak("Now we're not blocking the main thread.", routes="mic", background=True)
result = task.result(timeout=10)

Push-To-Talk Hooks

Playback hooks run after audio and routes are ready and after playback completes. They are useful for pressing and releasing push-to-talk around virtual mic playback.

from voice_conductor import PlaybackHooks, TTSManager

tts = TTSManager()

tts.speak(
    "Check out the playback hooks.",
    routes="mic",
    hooks=PlaybackHooks(
        on_audio_ready=lambda event: press_push_to_talk(),
        on_playback_complete=lambda event: release_push_to_talk(),
    ),
)

Documentation

• docs/general.md: architecture, settings, providers, and shared types.
• docs/mic-setup.md: virtual microphone setup and troubleshooting.
• examples/: small runnable examples.

Development

Install the project locally:

pip install -e ".[kokoro]"

Run tests:

pytest

Run a focused test file:

pytest tests/test_manager.py