I went through the process of setting up whisper.cpp on the DGX Spark and I wrote a python client to record audio → send to spark → receive text → save text to clipboard.
It works quite well and is very fast, so I thought I would share here.
The python script is fairly small, and the whisper.cpp setup is super easy, so this should be adaptable for any speech to text projects you have in mind.
I tested with a Mac laptop for the client and my DGX Spark for the server side.