Many thanks to the dev for this awesome repo! It's the best that I've tested.
I understand that hallucinations are intrinsic to the model, and it's not the dev's fault. But I was wondering if there is any way to reduce hallucinations?
Like output with phrases missing and strange noises in their place, or many little low, disruptive noises distributed along the output (are those artifacts from the silences in between chunks?).
I'm using the settings off the box.
Thanks!!
Many thanks to the dev for this awesome repo! It's the best that I've tested.
I understand that hallucinations are intrinsic to the model, and it's not the dev's fault. But I was wondering if there is any way to reduce hallucinations?
Like output with phrases missing and strange noises in their place, or many little low, disruptive noises distributed along the output (are those artifacts from the silences in between chunks?).
I'm using the settings off the box.
Thanks!!