VoiceStudio Default TTS Engine and How to Change It

VoiceStudio ships with OmniVoice as the default text-to-speech engine, utilizing the OmniVoiceBackend class (ID "omnivoice"), and you can switch to alternative engines like VoxCPM2 or IndexTTS by setting the OMNIVOICE_TTS_BACKEND environment variable or using the Settings UI.

VoiceStudio is an open-source AI voice generation platform that supports multiple TTS backends through a unified adapter pattern. The default engine configuration is defined in the backend/services/tts_backend.py module, where the system reads environment variables to determine which backend adapter to instantiate at runtime. Understanding this configuration mechanism allows you to optimize voice quality, latency, and resource usage by selecting the most appropriate engine for your hardware.

What Is the Default TTS Engine in VoiceStudio?

VoiceStudio uses OmniVoice as its default TTS engine. According to the source code in backend/services/tts_backend.py, the system checks the OMNIVOICE_TTS_BACKEND environment variable at startup, defaulting to the string "omnivoice" when the variable is unset (around line 13). This ID maps to the OmniVoiceBackend class, which implements the standard TTS interface for voice generation, model loading, and audio synthesis.

The current active engine ID can be retrieved programmatically using the active_backend_id() function:

from services.tts_backend import active_backend_id

# Returns "omnivoice" by default, or the custom ID you've configured

current_engine = active_backend_id()
print(f"Active TTS engine: {current_engine}")

How to Change the TTS Engine in VoiceStudio

You can switch between available TTS engines using two methods: environment variable configuration for headless deployments or the Settings UI for interactive use.

Method 1: Configure the OMNIVOICE_TTS_BACKEND Environment Variable

The most direct way to change engines is by setting the OMNIVOICE_TTS_BACKEND environment variable to a valid backend ID before starting VoiceStudio. Available engine IDs include "omnivoice", "voxcpm2", and "indextts".

Bash/Linux:

export OMNIVOICE_TTS_BACKEND=voxcpm2

# Now start VoiceStudio

python main.py

Windows PowerShell:

$env:OMNIVOICE_TTS_BACKEND="indextts"

# Now start VoiceStudio

python main.py

After setting this variable, the get_active_tts_backend() function will instantiate the corresponding backend class instead of the default OmniVoiceBackend.

Method 2: Use the Settings UI

For desktop or web interface users, navigate to Settings → Engine → TTS Engine. The dropdown menu displays all registered backends. Selecting an alternative engine updates the configuration and triggers a backend reload. The UI effectively writes the selected ID to the same environment variable or internal configuration store that active_backend_id() reads from.

Backend Implementation and Caching

The engine switching mechanism is implemented in backend/services/tts_backend.py (around lines 2696–2700). When you change engines, VoiceStudio performs the following sequence:

1. Unload: Calls unload() on the current backend instance to free VRAM and terminate sidecar processes.
2. Reset: Clears the cached backend instance via reset_active_backend().
3. Instantiate: Creates a fresh instance of the newly selected backend class when get_active_tts_backend() is next called.

This ensures clean resource management and prevents memory leaks when switching between resource-intensive models.

Practical Code Examples

Get the Current Backend Instance

from services.tts_backend import get_active_tts_backend

# Returns an initialized backend object (OmniVoiceBackend by default)

tts = get_active_tts_backend()
audio_buffer = tts.generate("Hello, world!")

Programmatically Switch Engines

import os
from services.tts_backend import reset_active_backend, get_active_tts_backend

# Change engine at runtime

os.environ["OMNIVOICE_TTS_BACKEND"] = "indextts"

# Clear the cached instance to force recreation

reset_active_backend()

# Verify the switch

new_backend = get_active_tts_backend()
print(new_backend.id)  # Output: indextts

List Available Engine IDs

While the specific registry lookup depends on the backend definitions in backend/services/tts_backend.py, you can typically discover available engines by inspecting the backend registry or checking the source for classes extending the base backend interface (such as VoxCPM2Backend and IndexTTSBackend).

Summary

• Default Engine: OmniVoice (OmniVoiceBackend, ID "omnivoice") configured via OMNIVOICE_TTS_BACKEND environment variable.
• Configuration Location: backend/services/tts_backend.py (line 13 for defaults, lines 2696–2700 for active backend management).
• Key Functions: active_backend_id() returns the current engine, get_active_tts_backend() returns the instance, and reset_active_backend() clears the cache when switching.
• Switching Methods: Set OMNIVOICE_TTS_BACKEND to "voxcpm2", "indextts", or other registered IDs, or use the Settings UI dropdown.
• Resource Management: The system automatically unloads previous backends and frees VRAM when you switch engines.

Frequently Asked Questions

What is the default TTS engine ID in VoiceStudio?

The default TTS engine ID is "omnivoice", which corresponds to the OmniVoiceBackend class. This is defined in backend/services/tts_backend.py where the OMNIVOICE_TTS_BACKEND environment variable falls back to "omnivoice" when not explicitly set.

How do I switch to VoxCPM2 or IndexTTS?

Set the OMNIVOICE_TTS_BACKEND environment variable to "voxcpm2" for the VoxCPM2 engine or "indextts" for the IndexTTS engine, then restart VoiceStudio. Alternatively, select these engines from the Settings → Engine → TTS Engine dropdown in the UI.

Do I need to restart VoiceStudio after changing the engine?

If you change the engine via the environment variable, you must restart the application for active_backend_id() to read the new value. If you use the Settings UI, the application typically handles the transition automatically by calling reset_active_backend() and reloading the backend without requiring a full restart.

Where is the active TTS backend cached in the code?

The active backend instance is cached in the tts_backend.py module (around lines 2696–2700). The get_active_tts_backend() function checks this cache, and reset_active_backend() clears it, forcing the creation of a new instance with the updated configuration.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}