Voicebox Studio

Voicebox: The Open-Source Voice Cloning Studio

Voicebox is a powerful, local-first alternative to services like ElevenLabs, designed for high-fidelity voice cloning and speech synthesis. It functions as a comprehensive creative suite, allowing users to clone voices from seconds of audio, generate speech in 23 languages, and orchestrate complex audio projects via a multi-track timeline—all while running entirely on your own hardware.

Key Capabilities

1. Multi-Engine Synthesis Voicebox integrates five distinct Text-to-Speech (TTS) engines, allowing users to choose the best tool for the task:

• Qwen3-TTS: High-quality multilingual cloning with support for delivery instructions (e.g., "whisper").
• LuxTTS: A lightweight, ultra-fast engine optimized for 48kHz CPU generation.
• Chatterbox (Multilingual & Turbo): Offers the broadest language support and paralinguistic tags for expressive speech (laughs, sighs, gasps).
• TADA: A speech-language model designed for long-form, coherent audio (up to 700s+).

2. Advanced Audio Post-Processing Powered by Spotify’s pedalboard library, Voicebox includes eight real-time effects (Pitch Shift, Reverb, Compression, etc.). Users can build custom presets or use built-in profiles like "Radio" or "Robotic" to polish their clones.

3. Professional Workflow Tools

• Unlimited Generation: Uses smart auto-chunking and crossfading to generate up to 50,000 characters without breaks.
• Stories Editor: A multi-track timeline editor for composing podcasts, conversations, and narratives with drag-and-drop ease.
• Version Control: Tracks "Takes" and "Effects versions" for every generation, ensuring the original clean output is always preserved.
• Async Queue: A non-blocking generation system that allows you to queue multiple tasks without crashing your GPU.

4. Voice & Model Management

• Profile Management: Create voice identities from recordings or files, supporting multi-sample inputs for higher cloning accuracy.
• Recording & Transcription: Built-in system audio capture and Whisper-powered transcription for seamless content creation.
• Hardware Efficiency: Local model management allows users to load/unload models to optimize VRAM usage.