I'm gonna do that then

I also took pi's good advice and removed unclear speech

I already submitted greg but once I have the improved version I'll submit that

true

Two, but epic.

W-what AI did you use?

VQGAN+CLIP

any prompts?

or...

no prompts I used images as the prompts

i got a question, i'm tranning my model right now with 500 files, and it's a bad tranning, the voice is clear and no noise what could be the problem i need help

looks like a transcription problem

thanks i got it,,,, i got alot of uh in my dataset

damn i'll never be abl to do a Dave Meltzer voice then 😔

you can still just cut out the speech disfluencies

but half of every sentence he says is an um or an uh

We need um and uh isolator then

Interesting

I assume it's quite easy to build into a script that just filters this stuff out

Just someone needs to be bothered enough by disfluencies to actually experiment a bit and write a script 😉

Hmm

Additionally, we present LibriStutter: a disfluency dataset based on the public LibriSpeech dataset with synthesized stutters. We also evaluate FluentNet on this dataset, showing the strong performance of our model versus a number of benchmark techniques.

What is this?

LibriStutter dataset is needed to train this model

Yeah there it is