search

haoheliu / voicefixer

General Speech Restoration
https://haoheliu.github.io/demopage-voicefixer/
MIT License
1.04k stars 133 forks source link

Artifacts on 's' sounds #38

Open JakeCasey opened 1 year ago

JakeCasey commented 1 year ago

Hello! Awesome project, and I totally understand that this isn't your main focus anymore, but I just love the results this gives over almost everything else I've tried for speech restoration.

However, I'm getting some interesting 's' sounds being dropped occasionally, and was wondering if there was perhaps a way of avoiding that, that you knew of?

UnvoiceFixed Voicefixed

Any ideas would be great, thanks!

adeelabbas commented 1 year ago

Hi, I am also seeing s sounds being dropped e.g. s sound in professionally. Any ideas how we can fix this?

RobertAgee commented 1 year ago

@JakeCasey your first audio is simply poorly EQd. A better mic and/or remastering your audio with something like remastermedia would work

Connum commented 1 year ago

@JakeCasey your first audio is simply poorly EQd. A better mic and/or remastering your audio with something like remastermedia would work

Erm, but isn't this all about restoring recordings with bad quality?!

RobertAgee commented 1 year ago

@JakeCasey your first audio is simply poorly EQd. A better mic and/or remastering your audio with something like remastermedia would work

Erm, but isn't this all about restoring recordings with bad quality?!

What if someone was voicefixing their podcast audio? 🤔 Or could it be possible someone (like yourself) completely tangential reads my comment and the solution applies to them? Many audio problems are basically bad setup or bad mixing, and there already exist better solutions for those scenarios. In any case, OPs audio doesn't need voice fixer, it needs EQ.

RobertAgee commented 1 year ago

@JakeCasey your first audio is simply poorly EQd. A better mic and/or remastering your audio with something like remastermedia would work

Erm, but isn't this all about restoring recordings with bad quality?!

What if someone was voicefixing their podcast audio? 🤔 Or could it be possible someone (like yourself) completely tangential reads my comment and the solution applies to them? Many audio problems are basically bad setup or bad mixing, and there already exist better solutions for those scenarios. In any case, OPs audio doesn't need voice fixer, it needs EQ.

@JakeCasey @Connum

Case in point: https://vocaroo.com/1k5uFdqY5jcy

And with a little more softness: https://vocaroo.com/1iQqmAWFyf0g

I don't know what OP is shooting for, but I did this in like 2 mins.

EmreOzkose commented 1 year ago

I am facing the same issue. Any improvement ?

madwurmz commented 1 year ago

Mode 2 would preserve that S sound, but it’s not always suitable. Sometimes voicefixer does mess up the results. @RobertAgee, your samples sound a bit dull but how did you upload them to vocaroo? I thought it only accepted microphone input. 🗣️ 🎵

RobertAgee commented 1 year ago

Mode 2 would preserve that S sound, but it’s not always suitable. Sometimes voicefixer does mess up the results. @RobertAgee, your samples sound a bit dull but how did you upload them to vocaroo? I thought it only accepted microphone input. 🗣️ 🎵

"Dull" aka soft sounding is subjective to whatever the end use case is. It sounds like he's making hypnotism audio, which may very well contain other soft sounding masking audio elements. I digress.

He just needs a de-esser to remove the harsh frequencies, and or a better mic that will capture a better spectrum. It sounds like he's using a phone or laptop mic that characterize the tinny sound of OPs audio.

Again, he's looking for remastering... And there's plenty of online solutions for that, even free ones.

There's a button that says "upload"... Click it

JakeCasey commented 1 year ago

@RobertAgee Wow thanks for the effort you've put in here! I agree there's probably some EQ that needs to happen so that voicefixer doesn't blow up the esses. I guess my biggest concern was I was hoping to keep the sharpness of the audio, while removing the occasional harsh s's. Compare your EQ'd versions with my voicefixed version, ignoring the S parts, and the audio is clearer, in my opinion. But I'm not an audio guy by any means, I'm just playing around with stuff.

For those following along at home, you can probably fix this to some degree with EQ before voicefixing.

RobertAgee commented 1 year ago

@RobertAgee Wow thanks for the effort you've put in here! I agree there's probably some EQ that needs to happen so that voicefixer doesn't blow up the esses. I guess my biggest concern was I was hoping to keep the sharpness of the audio, while removing the occasional harsh s's. Compare your EQ'd versions with my voicefixed version, ignoring the S parts, and the audio is clearer, in my opinion. But I'm not an audio guy by any means, I'm just playing around with stuff.

For those following along at home, you can probably fix this to some degree with EQ before voicefixing.

Again, it really depends on your end use case... so now knowing you want to maintain overall top end, besides the already mentioned solutions...

You could try adobe's speech enhancer online (it's free)

or use a de-esser plugin, which cleans up problematic sharp sounds in audio

Personally if I was you, I would try straight into Adobe first and see if that gives you what you want or close to it

With my ai vocals, I typically have sharp sounds on the high end, which i damp first with EQ then throw into Adobe that reEQs everything. Adobe can make things sound extra crisp.