Voxtral Codec : Combining Semantic VQ and Acoustic FSQ for Ultra-Low Bitrate Speech Generation (Voxtral TTS Backbone) - View it on GitHub