nemotron-labs-voicechat: Issues with the model's voice

I've been trying the Duplex voiceChat 11B model (via Docker) on an AWS H200 machine: https://github.com/NVIDIA-NeMo/Speech/tree/nemotron-labs-voicechat

It works ok, but there are some issues with the voice quality. It sometimes repeats itself. Is this expected? Is there any plans to fix/update it and commit a newer version?