LoudKit: Local TTS with Voice Cloning, 10-Language Support, and SDKs for Python, Swift, Go, Rust, and TypeScript

The open-source local TTS project LoudKit, featuring voice cloning and support for 10 languages, has been released. It is optimized based on Chatterbox, characterized by similar quality but smaller size and faster speed.

Key Features and Performance

• Multi-language SDKs: Provides SDKs for 5 languages including Python, Swift, Go, Rust, and TypeScript. Designed to generate identical voice tokens across each adapter, reducing quality variance by language.
• Two Variants: Includes the base model loudr-1 and loudr-1-turbo, which was trained to predict two audio tokens simultaneously by attaching additional heads to bottleneck points in the pipeline. The Turbo version is faster but may produce some artifacts.
• Voice Cloning: Works effectively with approximately 10 seconds of clean recording input; quality may degrade if there is background noise or long silences.

License and Data

• Code and model weights are distributed under the Apache 2.0 license.
• All included voice data was donated with consent or follows CC0/CC-BY licenses, with sources documented.
• The emotion axis feature is currently under development and is not included in this release.