"Iterations" when generating #251
Labels
No Label
bug
duplicate
enhancement
help wanted
insufficient info
invalid
news
not a bug
question
wontfix
No Milestone
No project
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: mrq/ai-voice-cloning#251
Loading…
Reference in New Issue
Block a user
No description provided.
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
I read in the wiki that Iterations are for improving the actual sound quality of your output audio... However, what is it actually doing to achieve that? Is is just noise reduction? I ask because I notice it skews to more tinny-sounding output.
From what I remember, the "iterations" in the web UI determines how many steps to run the outputted codes from the AR through the diffusion sampler (desu the names I went with are a bit confusing). I think I had some understanding that it would only really effect the actual sound quality, as it determines what goes into the final waveform, but that was with the old vocoder.
In my sparse generations after swapping the vocoder with BigVGAN, it seems that the "iterations" isn't all that necessary.
In other words, these days, as long as you are using the default setting for the vocoder (BigVGAN), you shouldn't need to stress about having as high of an "iterations" value as possible. I think 60 is what I just leave it to and it sounds decent enough compared to, if I used a different value for the same seed. VoiceFixer has been failing me more and more with some odd crackle at the end.