Open-source voice cloning app using Qwen3-TTS (github.com)

The newly launched open-source voice cloning app, Voicebox, enables users to synthesize voices and generate speech entirely locally on their machines, eliminating the reliance on cloud services. Powered by the advanced Qwen3-TTS model from Alibaba, Voicebox offers professional-grade tools similar to those found in digital audio workstations (DAWs). Users can clone voices with just a few seconds of audio, leveraging features like a multi-track timeline editor and batch generation for long-form content. The app maintains user privacy by storing all voice data on local machines, while its native performance on Apple Silicon provides significant speed advantages (4-5x faster inference). Voicebox is significant for the AI/ML community as it democratizes access to high-quality voice synthesis technology without the costs associated with subscription-based services. The app's API-first approach opens up possibilities for integration into various projects, including game dialogue systems, podcast production, and accessibility tools. Key technical innovations include smart caching for instant regeneration of audio prompts and extensive support for multiple languages. As an open-source platform, Voicebox also encourages collaboration and future developments, such as real-time synthesis and the ability to create new voices from text, positioning itself as a comprehensive solution for voice cloning and editing.