I was wondering about the audio in the podcast - it seems to sound distorted most of the time and the levels go up and down both between hosts and during clips. Do you know why that is? It seems weird that the audio would not be perfectly levelled when it is generated?
Hi Chris, it is a great point. I think partly because I trained the voice on speech that sounded natural rather than a standard newsreader. And I also think it is attempting to not sound generated (which would be more monotonous) and is missing the mark. Going to dive into more voice cloning this week, so will report back in the newsletter about how I get on.
Thanks for the comment and great to have a reader in Copenhagen!
Howzit from Copenhagen @Paul!
I was wondering about the audio in the podcast - it seems to sound distorted most of the time and the levels go up and down both between hosts and during clips. Do you know why that is? It seems weird that the audio would not be perfectly levelled when it is generated?
Hi Chris, it is a great point. I think partly because I trained the voice on speech that sounded natural rather than a standard newsreader. And I also think it is attempting to not sound generated (which would be more monotonous) and is missing the mark. Going to dive into more voice cloning this week, so will report back in the newsletter about how I get on.
Thanks for the comment and great to have a reader in Copenhagen!