Resemble AI Releases Chatterbox Multilingual v3

🎙 Resemble AI Releases Chatterbox Multilingual v3

Resemble AI has introduced Chatterbox Multilingual v3, an open-source TTS model based on Llama (0.5B parameters). It supports 25 languages and includes PerTh technology—neural watermarks resistant to compression and editing, which helps combat deepfakes.

🌍 The implementation of robust watermarking directly addresses regulatory requirements (e.g., the EU AI Act), simplifying the labeling of synthetic content. Utilizing the Llama architecture allows for the effective use of LLM advancements to achieve natural-sounding speech.

👤 You can use this powerful open-source model to create emotional voiceovers and perform voice cloning. Currently, the quality of the Russian language is in the mid-range segment (CER 1-5%).

Source 1: https://www.resemble.ai/resources/chatterbox-multilingual-v3-tts-with-embedded-watermarking-for-25-languages Source 2: https://huggingface.co/ResembleAI/chatterbox