Interesting.

Suno that describes the service as a collaboration between OpenAI’s ChatGPT (for lyric writing) and Suno’s music generation model

That would seem to suggest it’s two models and the audio is generated by the second.

Edit: from GitHub

Bark is fully generative text-to-audio model devolved for research and demo purposes. It follows a GPT style architecture similar to AudioLM and Vall-E and a quantized Audio representation from EnCodec