Help with boxy/muddy sounding voice

aram said:
I used to use Vegas too and I think if I went back to it, it would probably still be the video editor where I can work fastest (the way cutting works just made sense to me), but from version 12 onwards it just crashed on my system all the time (three different systems over the years).
Click to expand...

I've been using Vegas since 2004 ( back when it was made by Sonic Foundry and called "Video Vegas" ), and with Vegas Pro 16 I had reinstall a fresh copy of Windows 10 on both of my workstations in order to get VP 16 to run without problems. ( for whatever reason Vegas is very sensitive to Windows registry errors, I regularly use the free CCleaner registry cleaner to keep things running in top shape )
 
Ditto to all Dingo and yes to CC. I jump between Resolve and Vegas depending on the job. I also have Edius and used to use Premiere and years before that had two Discreet Logic edit* systems. Twenty years ago edit* was a great editor with a real time 2D and 3D hardware cards but was a shocker on audio and that's when I really started to use Soundforge and Vegas for audio. The programs were compiled on edit* then handed off to Soundforge/Vegas for all the audio QC and sweetening before network delivery. The fact that Soundforge can be invoked from the Vegas timeline and saved back into the timeline without ever leaving Vegas is a neat way to work. It's nice to be able to work down to audio sample levels in a video editor to remove a click or spike that's right in the middle of a frame.

Chris Young
 
aram ~

After reducing the boxiness sound as you put it to me it reveals that there is a certain mild harshness to his voice. A lot of how things sound I think has to do with taste. Exactly what the Vegas smooth/enhance is doing in tech terms I don't know. All I know is if you move to the negative it becomes a more 'damped' sound. You make a valid comment about hearing frequencies though. The older we get the more our high frequency hearing diminishes. When I was younger I had no trouble hearing the 15625Hz refresh rate of PAL CRT monitors. Struggling to do that these days but then I only have two SDI CRTs left. Both reference color monitors. A well respected custom amplifier builder in the UK once said to me "We know this high frequency loss affects older people because when we check the oldies setups many will have the treble settings turned down because as their hearing loses its elasticity higher frequencies begin to sound harsher and a little distorted. This may or may not be what you are perceiving as harshness. Get a decent set of speakers and a sine wave generator and wind it up until you hear harshness take a note of that frequency and then you know where your upper limit of hearing cannot be relied on for critical evaluation. Sad to say but it's true.

Finally as a last quick tweak taking on board what Dingo said about it still sounding a little flat and muddy in the mids and top end I had one more quick look and I think improved it by slapping an EssentialFX Male Voiceover plug onto the existing chain and giving it a bit of a tweak. I don't think I would be doing much more to it than that if it was my job as it always comes down to time vs budget and this result I know would not raise too many eyebrows in any of my intended audiences.

https://www.sendspace.com/pro/dl/a6lcdu

Chris Young

Boxiness + Vocal Strip.jpg
 
Going back to the original post - I prefer that to the now heavily processed version. Surely we are really talking about a very average recording with poor signal to noise. We can try to repair poor recordings and they're always a compromise. Getting it right in the first place is the key.
 
Agreed paulears, as I said in post #39. Better to get it right on the day and avoid heartache later.

Chris Young
 
Thanks for all your answers guys. Just got home from holidays and will have a look around and probably will have some more questions when I am done looking, tomorrow!
 
Bernie, I really liked that!

You clearly used some denoise (or at least a notch to get rid of that noise at about 960Hz)

I noticed that you softened a lot of the plosives by quite a bit and I think it works, beautifully. I had not noticed they were an issue until I listened (and looked at the spectrogram) of your version.

I saw you did high and low passes but your low pass starts only at 15K and seems relatively gentle. It worked for me!

Did you do any EQing besides the passes?
 
Chris, I think that person who told you about losing hearing elasticity nailed it. As a teacher I have to be in sometimes pretty noisy environments and there's a point when I start hearing these reverby distortions. It has also happened with music in places with a lot of reverb. It is as if some frequencies started accumulating to the point that the start sounding almost like the distortion that comes from clipping digital sound, even though the sound sources are as analogue as they can be (it happened to me once seeing Spencer Krug doing one of his piano only concerts in a church in London).

Gonna do a test at some point and see if I can identify those problem frequencies.

Thanks again to everyone, it is great to be learning from all of you!
 
Along with general hearing issues rooms can only handle so much energy. Once you pass the energy a room can handle you get distortions. Acoustic spaces like churches can easily overload with amplified sound, or even really loud acoustic instruments. Now on a strictly science level it may be that it's more your ears but there is also an issue that acoustically sound is pressure waves and if the sound gets too loud in an enclosed space the pressure never really gets back to atmosphere. Electrically this would be like adding a DC offset and it lowers the point where you will get noticeable distortion.
 
I'm glad you liked it Aram! I used spectral repair to kill that steady tone. I selected just it on the whole file using the rectangular selection tool and used attenuate in spectral repair to get it. I used the same selection technique to select from around 500 hz on down and reduced that whole range by 3 db using the gain module. I shifted the selection around until I found the muddiness. I did do some eq with some high and low pass and a small boost around 4K. I may also have done some spectral noise reduction on the over all file learned from the empty space at the head of the file. The plosives were addressed with the mouth declick module. To finish it off I selected some areas that had some interference sounding noises and used dialog isolate. This was all done in RX 6 on my small desk top speakers. I have another room all set up for just this type of work with RX7 but I wanted to see what I could do under less than ideal conditions. I've since tried some other techniques that also worked to an extent. The dewind module worked to thin out the boxiness a little more than the attempt I posted. The down side with that technique was it caused sibilance that wasn't easily addressed. The harshness you were hearing in some of the other efforts comes through on my small desk top speakers fairly obviously. I think certain frequencies were boosted too much and the result was greater clarity at the expense of a natural sound. I didn't find the mic to be that awful sounding and the levels were peaking at -9 in RX which to me is ideal. That's the level I strive to peak at when i'm doing location sound. That tone was the most annoying thing in your recording and spectral repair took care of it very easily. With better mic placement the Rode would definitely sound a bit better.
 
Last edited:
Thank you for the explanation!
I see you used mostly tools that I have (I am on RX5 standard, although got RX7 elements when it was on sale for 10 dollar and there some useful things there) and, more importantly, know how to use them!

I had not considered using de-click for plosives, that's clever. I was surprised to see that the plosives instead of being just in the low frequencies were pretty much all across the spectrum, which is quite similar to a click, so it totally makes sense.

Maybe one day I feel like I can splurge for advanced and get dewind and see what else it can do.

And yeah, I think the biggest issue with the recording in the end, bigger than the microphone itself, was the mic placement. Chin shadow lesson learnt (also it is not one that takes well to being put under clothes).
 
aram said:
Thank you for the explanation!
I see you used mostly tools that I have (I am on RX5 standard, although got RX7 elements when it was on sale for 10 dollar and there some useful things there) and, more importantly, know how to use them!

I had not considered using de-click for plosives, that's clever. I was surprised to see that the plosives instead of being just in the low frequencies were pretty much all across the spectrum, which is quite similar to a click, so it totally makes sense.

Maybe one day I feel like I can splurge for advanced and get dewind and see what else it can do.

And yeah, I think the biggest issue with the recording in the end, bigger than the microphone itself, was the mic placement. Chin shadow lesson learnt (also it is not one that takes well to being put under clothes).
Click to expand...
This was all done in RX6 Advanced in stand alone. The mouth declick is a bit different from declick and is designed to work on mouth noises. Declick does work on some mouth noises however. Dewind had the unwelcome side effect of increasing sibilance. You have a good ear and most of the tools you need to get good results. For inexpensive mics I also like the Oscar Sound Techs and have a few in each model. I mostly use COS 11Ds but when I need more mics the Oscars work well.
 
berniebeaudry said:
Here is another version I did. No EQ at all on this one. Thoughts? https://www.dropbox.com/s/13ek68tqtnj8qwu/boxiness_CU2.wav?dl=0
Click to expand...

That's really good (or at least it sounds great to my ears).

I can see you did not cut nearly as much with the low and high passes and that you did a bit more intense noise reduction, but somehow you managed to keep it natural sounding. I've just spent 30 minutes tweaking the noise reduction parameters to see if I could approximate your results and failed.

I assume you did multiple passes with a maybe 5 or so dB decrease in noise to such transparent results. Any other tip?
 
aram said:
That's really good (or at least it sounds great to my ears).

I can see you did not cut nearly as much with the low and high passes and that you did a bit more intense noise reduction, but somehow you managed to keep it natural sounding. I've just spent 30 minutes tweaking the noise reduction parameters to see if I could approximate your results and failed.

I assume you did multiple passes with a maybe 5 or so dB decrease in noise to such transparent results. Any other tip?
Click to expand...

Glad you liked it! Actually I didn't do any high or low pass this time, nor did I boost any frequencies. I used Spectral Repair to get rid of that tonal hum, then I selected from around 500 Hz down to the bottom with the rectangular selection tool across the whole file. Then I used the Gain module to reduce it by 4 db. I did 3 db in the first example. I followed that with Mouth DeClick. Then I learned the noise in Spectral de Noise from the front of the file and did a fairly big reduction in one pass with that. Reducing the noise revealed a slight smearing of one word which I think was a subtle mic rub on fabric. I used Dialogue Isolate on just that area to take care of that. Less is often more and trust your ears. I practice all the time with my own files and with others that people ask for help on. Your original file sounds far better than some of the terrible recordings I've taken a crack at. Yet, from your file I learned some new approaches that I will be able to use in the future. Finding the right order of processes and not over doing them I think is key to making things sound as natural as possible. Multiple light passes is a good approach for some noise reduction but sometimes it doesn't work as well as you would hope. Spectral deNoise when used properly can often work better in one pass with higher amounts of reduction. I tried getting that hum plus the noise with noise reduction alone and no matter what I did that tone wouldn't completely go away. Spectral repair was the best tool to use on that by far, then a separate pass of noise reduction could really be effective.
 
Last edited:
berniebeaudry said:
Glad you liked it! Actually I didn't do any high or low pass this time, nor did I boost any frequencies. I used Spectral Repair to get rid of that tonal hum, then I selected from around 500 Hz down to the bottom with the rectangular selection tool across the whole file. Then I used the Gain module to reduce it by 4 db. I did 3 db in the first example. I followed that with Mouth DeClick. Then I learned the noise in Spectral de Noise from the front of the file and did a fairly big reduction in one pass with that. Reducing the noise revealed a slight smearing of one word which I think was a subtle mic rub on fabric. I used Dialogue Isolate on just that area to take care of that. Less is often more and trust your ears. I practice all the time with my own files and with others that people ask for help on. Your original file sounds far better than some of the terrible recordings I've taken a crack at. Yet, from your file I learned some new approaches that I will be able to use in the future. Finding the right order of processes and not over doing them I think is key to making things sound as natural as possible. Multiple light passes is a good approach for some noise reduction but sometimes it doesn't work as well as you would hope. Spectral deNoise when used properly can often work better in one pass with higher amounts of reduction. I tried getting that hum plus the noise with noise reduction alone and no matter what I did that tone wouldn't completely go away. Spectral repair was the best tool to use on that by far, then a separate pass of noise reduction could really be effective.
Click to expand...

It's interesting to me that you did not cut. I looked at the spectrum and I guess the denoise process itself took off most of that below 100Hz and over 15KHz energy. RX5 standar does not have the gain module, but RX7 elements does, so that's something I'll give a go this weekend (I am a teacher and work is brutal during the week). I have just found out about Dialogue isolation... Ah, would it not be nice to be rich and happily pay for RX advance.
I think the only "but" I would put to this is mostly my fault. The s at the end of contactless is pretty brutal! :D
 
aram said:
It's interesting to me that you did not cut. I looked at the spectrum and I guess the denoise process itself took off most of that below 100Hz and over 15KHz energy. RX5 standar does not have the gain module, but RX7 elements does, so that's something I'll give a go this weekend (I am a teacher and work is brutal during the week). I have just found out about Dialogue isolation... Ah, would it not be nice to be rich and happily pay for RX advance.
I think the only "but" I would put to this is mostly my fault. The s at the end of contactless is pretty brutal! :D
Click to expand...
Ok, I accept that challenge LOL. Here's a new version that addresses the s that is bugging you. Although to my ear it wasn't that bad, but you recall the conversation earlier on about taste. De Ess wasn't touching it without problem. I used De Rustle on it and one other area to improve the sound of it. https://www.dropbox.com/s/ykjnnxp5m12e8h5/boxiness_CU3.wav?dl=0
 
Sorry about the long absence. "In real life" I am a teacher and life goes to hell once the term starts. This is the first time in three weeks that I have a few hours for myself without being extremely tired.

I listened to it a few times and I could tell the main offender "contactlessssss" was different, but not how exactly. l did not like it better (or worse). So... I opened RX7 Advances (Izotope can thank you for getting my money, I had been pretty happy with RX5 standard for years, but hearing what you have done, I got greedy for more) and had a bit of a play myself. I can get the sibilance in the later syllables to sound fine, but that contactlessssssss is too much.

Thank you once more, I have learned a lot from this thread and next time around I am sure I'll do a better job. Now, time to write another script!
 
Actually, one more question. Looking at the spectral view of the latest version I noticed energy cuts before and after the sibilant areas, but the reduction of the main "body" in the s sound was limited. Was this done by design, because you like it like that (it does shorten the sibilant sounds a bit, making them less obvious) or was that the result of some specific tool you used this time around?

I quite liked the result of using the spectral de-esser with a big tilt to brown noise and very high spectral shaping (used it on the CU2 version).
 
You must log in or register to reply here.
Back
Top