May 29, 11:38 PM

Resemble AI Launches Chatterbox Open-Source TTS Model With 0.5B Llama Backbone, Outperforming ElevenLabs Conversational AI 2.0

Resemble AI has released Chatterbox, an open-source text-to-speech (TTS) model that has been preferred over ElevenLabs in A/B testing, with a 38.75% preference compared to ElevenLabs' 27.5%. Chatterbox is built on a 0.5 billion parameter Llama backbone and trained on over 500,000 hours of clean audio data. It features an exaggeration dial for expressive speech and supports zero-shot voice cloning, allowing it to replicate voices without prior training. The model is emotionally aware, capable of expressing various accents, tones, and styles, and is available under the Apache 2.0 license. It has quickly gained popularity, becoming one of the fastest-growing open-source TTS models with over 2,000 stars on GitHub within 24 hours of release. The model is accessible on Hugging Face, where users can test demos and integrate it into applications. Meanwhile, ElevenLabs has introduced Conversational AI 2.0, a voice assistant platform that supports multimodal interaction with voice and text simultaneously, featuring advanced turn-taking, language switching, multicharacter mode, batch calls, and built-in retrieval-augmented generation (RAG). ElevenLabs' platform is enterprise-ready with HIPAA compliance and EU data residency. Both companies are advancing AI voice technology, with Resemble AI focusing on open-source expressive TTS and ElevenLabs enhancing conversational AI capabilities.

#Resemble AI #Chatterbox #ElevenLabs #Llama #Apache #GitHub #Hugging Face #Conversational AI

Written with ChatGPT (GPT-4).

Sources

Additional media

Image #1 for story resemble-ai-launches-chatterbox-open-source-tts-model-0-5b-llama-backbone-ai-2-0-82c60c6b