How to remove raspy sounds (vocal fry, crackle) from voice

Hello, I’m not sure this is the right forum category. I tried to contact Steinberg sales/support but the website only provided a link to this forum.

I’m not even sure what product I need, but I’m guess it might be something similar to SpectraLayers or maybe Nuendo or maybe one or a combination of other sound or voice editing products. I can copy this to the other categories if necessary.

So on to the problem:

I’m working on an audiobook project recorded by the author.

Unfortunately, she’s not a professional VO artist and she speaks with a somewhat raspy voice.

I’ve gone round in circles trying to identify what is causing this “gravelly”, raspy sound or “crackle” and to remove it.

I have RX8 Standard. And none of the modules seem to have any effect on reducing or eliminating this noise. I also can’t seem to identify which part of the audio spectrum it’s originating from.

Please could you provide some guidance on how to go about editing the audio file.
What modules/tools would I need to use? How can I identify the noise in the spectrum editor? And is there a way I can create a filter that automates the process of isolating and removing the noise for the entire file?

BTW I have similar problems for most other voiceover recordings, it’s not a unique problem to this particular artist. I think some people call it vocal fry and it’s supposedly desirable in certain kinds of jazz songs I suppose (like Satchmo).

But these aren’t songs. They are audiobooks, and the raspy sounds become very irritating after even a few minutes - a bit like when someone has phlegm in their throat and won’t bother to clear it away.

I’m already using Izotope RX8 and have tried to get help from their support, but it doesn’t seem that any of their modules can deal with the strange voice noises.

I’ve tried declicking, dialogue isolate, spectral denoise, voice denoise and nothing seems to reduce or eliminate the raspiness.

Is this something that your module(s) can deal with?

Even if you don’t know how to fix these issues yourself, it would help if you could identify what they are. At least then I can search for filters designed to deal with them.

It’s actually such a common noise problem that I’m surprised no-one else seems to have noticed it. I’ve even heard it what should be professionally-edited productions, e.g. Audible or in movies.

I thought these raspy sounds (in this particular talent’s voice) were around the 1.5KHz mark (+/- 500Hz) but another colleague of mine seems to think that they’re chunks well below the 1KHz range (a big blotch of sound from about 50Hz to maybe 800Hz or so)…

Thanks in advance. I look forward to your reply.

Gary

PS I’m new to this forum - can anyone also explain to me how to upload a sample file?

I guess there’s no chance of convincing the author to step aside and let somebody else do the VO? That might be easier than having the mastering engineer kill himself trying to correct a fundamentally flawed recording. Sometimes you just have to give up and start over.

That said, it’s possible that simple EQing would make the rasping problem at least less noticeable. I’ve done a remarkable lot with my Voxengo para EQ, but I don’t know how well it would work against your problem.

Addendum: SL, unlike RX, doesn’t allow you to review FX settings in real time. To me, that would be a real drag if I were trying to dial in precise EQ settings. Wavelab Cast, OTOH, allows you to review FX settings in real time. I’m sure that applies to all versions of WL.

Thanks Poinzy. Whether I get someone else to do the VO or not, I still have the same problem. It seems that nearly everyone has some kind of raspiness in their voice and it can be fairly distracting.

So I’m looking for a way I can remove it or make it less noticeable in general. :slight_smile:

I’ll look at WaveLab Cast, thanks. I’ve gone round the various software options on the Steinberg website, but I still have no idea what each of the versions or modules do. Is there a page that actually lists each of the filters or effects in each version of the software packages and what they do? (And when it says “limited”, in what way are they limited?)

Voxengo EQ also seems really great, btw!

Can you post a short sample so to have a better understanding of the problem and see if that’s something SL could fix ?

Thanks, Robin. I couldn’t find a way to upload a sample on this forum. And then I noticed the “upload” icon, but it wouldn’t let me upload wav files. There’s no point in uploading a lossy file coz then you can’t clearly identify or isolate the noise. Oh and I can’t include links either it seems! :o

So I’ve saved it online at orish-dot-com in the /vo folder and the filenames are lynnsample.wav and lynnwrote.wav.

The first file includes a weird noise that seems to come from somewhere inside Lynn’s body, maybe her shoulder bones rubbing against each other. I can more or less isolate and remove these noises manually.

But the real problem is the raspy sound that you will hear clearly in the second file where she says “wrote”.

@garyorish So… on one hand, the bad news is that there no fully automated process that can clean your problem, no matter the software (as far as I know). The good news though, is that you can definitely clean such vocal fry occurences with SpectraLayers tools.

Here’s the problematic area:


Between 0Hz and 1KHz, you can sometime see noise between the voice harmonics, this is the unpleasant sound you want to get rid of.

The solution is to use the Harmonics Selection tool, set the thickness to something very thin (1 pixel), point to the fundamental frequencies and swipe left/right to automatically select the rest of the frequencies:

Then click Select > Invert Selection Frequencies (shortcut Ctrl+Shift+I or Cmd+Shift+I on mac):

Then use the Rectangular Selection tool (shortcut M) in “Intersect Mode” (hold Shift+Alt) and only retain the part between 0Hz and 1Khz:

And press Delete (or click Edit > Delete):

You can then safely deselect (shortcut Ctrl+D or Cmd+D on mac) and listen to your cleaned frequencies:

And here’s the cleaned result, which you can compare to the original lynnsample.wav.
It sure takes manual work, but once you have your shortcut ready it can be done pretty fast. I’ve processed the 30 seconds sample in 3 minutes.

3 Likes

Hey Robin, thanks for that. I see you put a lot of effort in finding a solution.

A few things.

I did notice that the sound noticeably receded at the point where the raspiness was removed (diminished). I suppose one would then have to add a gain to the specific region that was filtered, right? (Or maybe some kind of compression afterwards to keep the audio level throughout?)

You say that I can create a shortcut. That would be very useful. Do you mean that a sequence of steps can be programmed to select, invert, intersect, filter, etc.? So all I’d need to do is to go through the audio manually and select the bits that need to be processed and then run the “macro” (script, program)? Is this part of the module?

Finally, how do I buy the Harmonics Selection Tool? All I found was the help page for this module, but there’s no listing in the products section to actually purchase it. I see a list of software products, but no mention of this module. Is it like a VST that I can add into my audio editor (Audacity? Sound Forge? RX?)

Thanks again.
Keep well,
Gary

@garyorish To make it sound similar to the original recording in term of EQ, I then suggest you use another process introduced with SpectraLayers 8, EQ Match.

There’s no macro system yet in SL, I was refering to shortcuts to each tool/action (Harmonics Selection tool, Invert Selection Frequencies, Rectangular Selection, Delete, Deselect). This will save you time not searching them in the UI.

As you can see in this edition comparison page (or this page from the doc) the Harmonics Selection tools (as well as the EQ Match process) requires the Pro edition of SpectraLayers.

Robin,
Anyway you could make this into a quick tutorial? These are the more advanced instructions that I would love to see.

Thx!
B

1 Like

Thanks. I’ll download the trial and play with it for a bit.

I did look under edition comparison, but I didn’t know what I was looking for - their was no mention of Harmonics Selection tool - is it a filter, a process, surely it’s not in the section marked “selections”? (anyway I didn’t see it there either, and no explanation what the various features do, so it’s obviously designed for experts who already know about each of the components in the various packages).

I did see it the docs page, that had a better list of what’s on offer. But it seems expensive to have to buy a whole software package (and the most expensive pro version at that) just to use the one module after all…

As for EQ Match, that will lift up the parts of the signal that were reduced in volume during the harmonics selection (or removal) tool, right?

It all seems kind of complicated, and I’m wondering whether this is the right approach… :slight_smile:

I’ll be happy to minimize the raspiness for you for a fee. Private message me.

how to send messages?

Well, I’m looking for a solution for all the voiceover editing work that I do, because so many people (even the professional vo artists) tend to have this aspect to their voice and it can become jarring for long works like audiobooks or documentaries (esp. when there’s no musical soundtrack to distract from it).

But if you’re not expensive then I wouldn’t mind passing on just the final editing for you to do (manually going through and identifying/removing the raspiness). There will be a lot of it.

:slight_smile:

@Robin_Lobel … thank you so much for the vocal fry workflow - it is truly amazing.

I do my own VOs and, unfortunately, my voice suffers from vocal fry particularly on some vowels, as well as mouth clicks and so on. It usually takes a long time in Sound Forge Pro to edit the best parts from the various takes and then using the draw tool to reshape the sine wave to reduce/eliminate the raspy bits, or to just delete a sine wave or two, or to re-record the vowel/word and edit it into edited ‘read’. Now, with the SpectraLayers workflow (which I tested today), it is going to be so much easier and faster to remove vocal fry. Thank you again.

As much as I like iZotope’s products, I don’t think that there is anything similar to remove vocal fry, even in RX Adv. Some years ago, in an iZotope survey I received from them, I suggested the removal of vocal fry as a desirable feature for a future RX release, but I haven’t seen anything along those lines. I think that the big difference between SpectraLayers Pro and RX Adv is that RX is heavily weighted on automated processes whereas SpectraLayers is far more ‘surgical’ as one prominent contributor to Vegas Pro’s forum well described as its biggest advantage years ago on the forum.

And for me, SpectraLayers Pro’s Clone Tool - with its adjustable settings - has been an invaluable tool for me for many years especially for replacing low to mild wind noise (usually if under 500 Hz) in background location audio with other areas of the recording that are clear of wind noise.

1 Like