This seems to work well for certain voices and recordings and quite badly for others.

Is there any way one could easily train it on my own voice to make sure it isolates my or any other trained voice from noisy environments?