Integration of the Open-Source Sesame CSM-1B Conversational Speech

[rodrigoGA]

It would be necessary to add support for the open-source Sesame CSM-1B model within NVIDIA NeMo’s TTS framework. The CSM-1B model uses a Llama backbone paired with an audio decoder to generate RVQ audio codes from text and audio inputs, making it highly effective for creating natural, conversational speech.