Audio Finishing for Best, Most Even Sound

Hello,

I make Youtube gaming let’s plays, and this is not the first time I have asked for help on the Audacity boards, about specific things, but this is my first time on the podcasting subforum, as gaming LPs are essentially podcasts and have from an audio perspective the same elements.

First off, I am not asking for a miracle or how to have “radio announcer” voice. I am simply looking for the best way to present my own voice in a natural way.

My current post-processing set up is: 1) Noise removal highlighting an area without voice. As far as I can tell the settings I use here does not harm my audio. 2) Next, I use the exemplary de-clicker tool by Paul L which I love. Spent a lot of time tweaking it to suit my needs. I’m sure it distorts the sound in some way but it still sounds great to me in Audacity up to this point. Both of these things are necessary for me. I unfortunately have computer fan noise and mouth clicks that I cannot otherwise remove.

At this point I’m not sure of what I should be doing. My main goal that I am trying to achieve is loudness, evenness, and clarity. Speaking for 20-25 minutes at a time there is no way to prevent myself from going soft in some parts and loud in others, so a 3) compressor is as far as I know necessary. I also 4) normalize, and 5) do a very minor EQ where I raise the bass by about 3db and lower the high treble by about the same.

My mic gain is set pretty low so my initial waveform is pretty tiny. On the scale from 1 to -1 it honestly looks to go from about 0.1 to -0.1. Youtube tutorials would have me normalize, compress, equalize, and then normalize again. This is successful in that my volume is even and loud and clear. The problem is that it also sounds a bit mechanical and there is a bit of a metallic hum occasionally at the beginning of my words, almost a high buzz.

A good example of audio processed in that way is here: https://youtu.be/NwNfy5HvFRw . A bit tinny to my ears.

So then I tried compressing first, then EQ, and finally normalizing only once. This audio sounds better to my ears, but 70% of it is too soft, due to a few high peaks. I rectify this by raising the loudness of the whole thing by about 3-5 db and then lowering the loudness in small envelopes around the peaks. Still, it never is as loud and clear as the twice normalized version, plus it takes me twice as long to edit, but there is never the metallic sounding him in some words like the twice normalized audio.

A good example of this set up is here: https://youtu.be/zP4N7MSKdO0 . The vocals are obviously quieter and have less punch.

My overall questions boil down to:

  1. Does the effect of a compressor change whether it’s done before or after normalization? I would presume yes, as louder audio would be more compressed, or to be more clear, assuming the same threshold, more sound will lie above it in the normalized version. Adjusting the threshold specifically for each video though would be tedious.
  2. I have tried compressor settings from 1:3.5 to 1:5. From a results standpoint where you want to limit loud noises and create more even volume it would make sense that the higher, the better, but is that true?
  3. Would a hard limiter perhaps be a better choice?
  4. Would a leveler be superior to a normalizer for my purposes?
  5. I have heard it said that boosting treble adds more warmth and clarity to voice, but wouldn’t it also add to harsh ss’s (sibilance) and the sharpness of my voice, which isn’t very melodic to begin with.
  6. Is it better to turn up gain on mic so it has to be normalized less? Most advice I have seen is to keep gain on mic way down to avoid picking up ambient noise, which I don’t have a large problem with my current low gain settings. But I feel like normalizing just magnifies the soft sounds (like my breath) just as much as the loud sounds (my voice).

And because I’m not an audio professional, if you do check out the above links, I’ll take any other thoughts or advice that you have as well.

Thank you!

-Marcus

I like the first one. I don’t think it’s tinny at all. I think it’s more mellow and easy on the ears. Are you sure you didn’t get those labels backwards? The second one has a stiffer and harder sound in my opinion.

You already know one straightforward way to improve the quality of the work is eliminate the fan noise and reduce the noise reduction. Reducing noise reduction will help reduce the words “tinkly, honky, brittle, and mechanical.” Excess Noise Reduction can do all those things.

A noisy studio restricts the ability to set the microphone. Normally, you place the talent so it’s as loud and close to the microphone as possible without stressing the performer in any way or creating breathing/wind swishing noises, pops and clicks. There’s a whole thing about microphone placement and the use of pop and blast filters, etc.

This one is wrong.

http://kozco.com/tech/audacity/clips/PPopping.mp3

The louder you are and more you turn the microphone down, the less the sound channel noise (fffffffffffffffffff) and the less noise reduction you need in post production. In a good room with good equipment (and noiseless computer), I was able to create an AudioBook quality clip with no processing other than volume change. I wanted to see if I could do it.

There is a restriction for getting the sound too loud, but modern microphone systems tend to record quiet to get around that. This design “feature” usually makes the ffffffffff noise worse. Pay attention to the overload indicator on your microphone system and make sure the Audacity sound meters never go all the way up and turn red.

This is what a perfect recording looks like.

Nobody will come out with a stick if you don’t make that, but you shouldn’t get too far away. Certainly no louder.

~~

Enter the computer noise. It is no longer just convenient to make your live voice loud and clear, it’s required. You need to get your voice louder than both the tiny fffffffff sound of the microphone system and the background whirr and rumble of the fans. One AudioBook instructional video has you putting the computer outside and run extension cables into your studio.

I designed an odd way of using the compressor. The metaphor is:

  1. Adjust the work somewhere predictable.
  2. Apply the compressor.
  3. return the work to a predictable level.

This is how that shakes out using AudioBook settings:

Audio Compressor
– Select the whole clip or show by clicking just above MUTE.
– Effect > Normalize: [X]Remove DC, [X]Normalize to -3.2 > OK
– Effect > Compressor: Thresh -20, Floor -40, Ratio 2:1, Attack 0.2, Release 1.0, > OK
– Effect > Normalize: [X]Remove DC, [X]Normalize to -3.2 > OK

That -3.2 number is particular to AudioBooks, but it’s not a dreadful volume for spoken word. You don’t have to use it. It can be higher, although I’d stay away from -0-.

You would think that’s a lot of work just to apply the compressor, but if you do that, the only character of the work that changes is the compression. If you don’t like the compressed sound, UNDO it and apply a new value. The only thing you change is Ratio. All the other values and settings in all three steps stays the same.

The overall sound will not wander; the only thing that happens is the work will now have your new compression values. Keep UNDOing and changing Ratio until you get the sound you want.

If you don’t do that, the compression sound quality will change with the loudness of the live performance. You can compensate for that by getting good at the other compression adjustments. Once you do change the compression, the finished volume may be different, too. You can fight with that if you want. I’ll use that three-step process.

And way down at the bottom of the list is the mistake. There is a theatrical mistake in the first clip. You picked a wrong title or put a colon in the wrong place? You do edit those out before submitting the final work, right?

Koz

Thank you for taking the time to post such a detailed reply! I will continue to look it over to make sure I’m getting it all, but the main thing I noticed was that you are in support of the double normalizing (before and after compression) and don’t believe it affects the audio negatively. That is not the conclusion I came to so it looks like my analysis was wrong somewhere else, but I’m glad to hear it.

I would love to have a situation with less fan noise and I’m working on purchasing a quieter computer since moving my computer is not really an option due to where I’m recording. There’s a lot of things less than ideal about my “studio space” (which is essentially an open closet with a desk in it, leading to a close computer and echo) but that’s not something I can change without considerable expense. I will try to lower noise removal settings without reintroducing the noise.

One follow up question regarding noise removal: is it better to have a longer sample or a shorter one? My current sample is about 3 seconds long.

It also appears that you endorse the notion of turning gain on the mic (and computer software) way down and just raising it in post. Compared to your perfect recording image, as I mentioned earlier, yours tends to oscillate from .5 to -.5. Mine is more like .1 to .-1 even when I’m loud (on your image below, look at from 14.5 seconds to about 14.6, that’s what my entire wave looks like pre-processing). So while in general you endorse a more quiet mic, in my case, seeing as it’s so much smaller than yours, would you recommend I turn the gain up?

You mention your compressor usage as “odd”, but they mirror what I see a lot of places. What is a typical or usual compressor usage? I have had better luck with the built-in Audacity compressor than the more fancy ones such as compress dynamics and SC4 which are both also in Audacity, but they both seem to make quiet noises like my breathing louder while the built-in compressor does not. Maybe I’m using them wrong. As for normalization, I typically normalize to -1 since that’s generally what you’re told to do for Youtube.

And I couldn’t really understand your last line: “And way down at the bottom of the list is the mistake. There is a theatrical mistake in the first clip. You picked a wrong title or put a colon in the wrong place? You do edit those out before submitting the final work, right?” Are you referring to the link to my Youtube video? If you are referencing the video title, it’s the name of the game and the faction within the game that I’m playing, along with the episode number and episode title. It’s a typical layout that is used commonly.

Thank you!

-Marcus

I would love to have a situation with less fan noise

There are things you can do with what you got.

Put the computer on several layers of carpet, towel or blanket to keep it from coupling noise into the floor. Wave your head around the computer and see where the most noise is coming from. Mine come from the power supply vent system in the rear. Place heavy blankets, carpet, etc. on the wall opposite that side of the computer. The object is to have fan noise exit the computer, cross the room once and stop dead. It’s tricky because you’re not allowed to block any air flow.

Float the microphone.

And etc. There’s no end of noise abatement tricks. I once got stuck recording a voice track in a large warehouse. I found the location with the least hum and air conditioning noises and set up shop. About a quarter of the way in, the laptop got hot and decided to cool itself. !@#$. I put it on a chair with heavy coats facing away from the performer. The room was big enough that the room reflections didn’t make it into the show.

That blue thing under the microphone, book and towel is a furniture moving pad. That keeps the desk from making or reflecting noise. Note there’s a double one on the conference room table in this shoot (the conference room itself is soundproofed).

And etc. Note that semi-circular cave thing around the right-hand microphone (the performance was double recorded). That was the performer’s portable sound shield. That’s not a dreadful thing to do, either and no, you don’t have to blow the big bucks. See: furniture moving pads.




My current sample is about 3 seconds long.

That’s probably fine. You should make surgically sure there’s only noise in the sample/profile. Do Not Move. Hold Your Breath. There was a recent poster who never quite got it that shuffling, chair squeaks, clothing noises, and arranging papers were poisoning the noise reduction profile.

you endorse the notion of turning gain on the mic (and computer software) way down

I don’t know. This is where you get to tell us what you’re using. Model numbers are good.

There is a theatrical mistake in the first clip.

At 2:33 in the first clip, you mess up a title or heading. “apparently, we’re not going to do that.”

Mistakes should never make it out the door.

Koz

Thanks for the info. I use a Blue Yeti USB microphone set in cardiod mode. It’s probably way below the kind of mics you work with, but my channel is not my income source and is a part time hobby (I don’t even sign up for ad revenue) so getting a traditional mic with an amp and all that is not in the cards. There are other USB mics that compete with mine, and while none are described as wholly superior, some claim to be more cardiod and resistant to noise (the Yeti has a reputation for picking up everything).

I have considered the sound shield. Like your photo, I have my mic on an arm and it hangs in front (and slightly to the side so that I can see the monitor) of me . The shield would be too heavy to mount so I would need to put it on the desk and so far the ones I have seen have been way too big for this. If someone made a half size one that just fits around the mic that would be great.

I am very careful to make no sound during the part of the recording I will use for noise removal base.

As for theatrical mistake - the nature of this form of entertainment, if you can call it that, and I presume a podcast is similar, is that you’d be spending 10x the length of the recording itself to wholly remove all verbal miscues and errors. Despite my channel being amateur I put a lot of work into improving it (hence why I’m here) but I’m not going to go overboard. There are other LPers with 100 times the viewers that I have that are less “polished” in presentation but are fun and interesting.

Thank you!

-Marcus

Blue Yeti USB microphone set in cardiod mode.

Then the settings may be pretty much take care of them selves. Turn everything all the way up. Yetis run low volume, so it’s difficult to overload either the microphone or the sound channel. I’m betting if you crank everything all the way up, you still don’t get anywhere near “ideal” volume in Audacity.

As a marginally silly experiment, intentionally overload the microphone system. Does the Yeti have an overload light? See what you have to do to cause Audacity’s recording meters to max out and turn red. I bet you have to scream into the microphone. That’s your maximum performance volume. Any vocal volume lower than that is fine. Do Not Ever blow into a microphone, but almost any sound is fair game.

Did we establish you were on Windows? Some Windows installs have a 20dB Microphone Boost setting. I would probably not set that. Those can be noisy. Manufacturers found it valuable to have that setting because if everybody is gun-shy about accidental overload (which is permanent and fatal), the system stops working. The microphone is low volume, the preamp is low volume, the analog to digital converter is low volume, etc.

Son of a gun. The performance is too low. We’re going to have to get volume boost from somewhere…

Draw a line from the Yeti through your head to the wall behind you. That’s generally what the Yeti picks up. Pile soundproofing (cardboard boxes, blankets, etc) back there. That should not be a blank, hard wall.

You don’t have to create a microphone sound cave. That’s just one option. Does the Yeti mount have any rubbery, gooshy or springiness to it? If it doesn’t, then the Yeti is picking up whatever the mount is sitting on. Extend that book and towel idea. I’ll bet significant chocolate you can get a vibration mount for a Yeti.

Or make one:
http://www.kozco.com/tech/pvcShockMount/shockmount.html

you’d be spending 10x the length of the recording itself

Theatrical Production uses 5X as a fuzzy rule. It’s rarely faster than that. So yes. You hit it.

Koz

These settings and adjustments are general recommendations, comments and observations, not laws. If you find different values or settings work, then go for it. For example, I can tell you exactly where to set the knobs and dials on a sound mixer to start out. No one is expecting them to stay there once the performance settles in.

We did miss a step with you because you’re so far along. Record and post a raw sound test.

http://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

I’m not interested in noise reduction, boosting or normalization. Press record, talk, press stop, export and post it. No corrections. Don’t “help” us.

Koz

Hello. I do have a shock mount which the mic is mounted on. The shock mount is mounted on the boom arm. It’s one of the star shaped ones. I also have a pop filter attached as well.

Wow, that’s so interesting, so even though the mic is cardioid, it records from directly behind itself. It’s not only facing a hard wall, but a corner as well. If I hang a blanket in that corner it would help with echo? That’s quite useful to know. (EDIT oh wait, you mean behind ME. Um, that’s a hard wall, but it’s slanted towards me so if sound waves move like a thrown ball the sound wouldn’t bounce back to me but to an angle, where I have a floor to ceiling, wall to wall bookcase, and the room has carpting. Now I’m wondering where the echo is coming from. I thought it was coming from the tiny alcove closet space behind the microphone).

I’ll post a sound test but just be warned there is going to be a lot of mouth clicks! I thank God I found that plugin by Paul L.

That’s funny what you say about the Yeti. Word on the street (and my own experience) is that’s it’s super sensitive, with people suggesting the rode podcaster and Audio Technica AT2020 as better for poscasting/LPing, even though they are cheaper mics that technically produce inferior sound (so the reviews/comparisons say).

-Marcus

Attached is the 20 second clip. The mouth clicks are killing me. I wouldn’t be able to do this without that plugin. Apples, dry mouth lozenges, spray, licorice, lemon water, nothing works but that plugin.

-Marcus

it records from directly behind itself.

No. It records directly behind you.

Draw a line from the Yeti through your head to the wall behind > you.

The recording pattern is a balloon and it’s recording you and behind both of your ears. It’s not a narrow laser beam.

There is a pattern called supercardioid or hypercardioid which does record a little directly behind itself, but the Yeti won’t do that.

The shock mount is mounted on the boom arm. It’s one of the star shaped ones. I also have a pop filter attached as well.

Then you win. That’s exactly what you’re supposed to be doing. I’m surprised you have mouth noises… We’ll see. You wouldn’t be the first performer dying of mouth noises nobody else could hear.

it’s super sensitive

Can you make a normal volume recording in Audacity according to that graphic? They mean it’s sensitive to the dog barking across the street. That usually means it’s receiving almost everything in front of it and not focusing on your face. That balloon sensitivity pattern isn’t standard. It can get broader and narrower, usually with price. Lower price microphones can have sloppy, broad patterns and receive from the sides, too. More expensive ones can have well-behaved, good-sounding narrow patterns and less likely to pick up the refrigerator, Air Canada jet going over, and Metrobus.

Attached is one of my microphones. It can switch patterns and it has two normal cardioid patterns, plus the others. The fourth one is HyperCardioid with that little sensitive zone hanging off the bottom. If you were directly behind that one, you could still be heard, but the front pattern is much sharper and less likely to pick up room noises.

Koz
Screen Shot 2016-04-12 at 19.59.56.png

Dueling posts. I’m listening to your sample.

Koz

It sounds like a machine shop back there. No, strike that. I know machine shops that don’t make that much noise.

Everything else you’re doing on that clip is fixable, most of it easily, except for that. I stopped trying when I got up to 18dB noise suppression Beyond that, the voice starts getting damaged almost no matter what we do.

By the way, you have two noises. I know you lump everything into “fan noise” but you have at least two powerful fans running slightly off-speed with each other. It takes a couple of passes to hear it.

mmmmmmmmMMMMMMMOOOOOOOAAAAAAAOOOOOOOMMMMMMMMmmmmmmmmmmm.

That’s why it’s so difficult to get rid of it. It changes. Audacity Noise Reduction doesn’t do changing sounds.

So without question start there.

The clip is about 15dB low. Sound doubles every 6, so you’re seriously low. Did you do that overload test? Can you make a recording that matches the graphic I sent? Even if you turn everything all the way up and yell?

The voice is clear with good diction, pronunciation, etc, etc. In My Opinion, it’s very slightly bright. I “turned the treble down” a little bit (4KHz equalization droop) and out the door. I don’t hear terrific mouth noises or anything that would demand heavy de-essing or sibilant processing. That’s up for other’s comment.

How are you listening? I have a newer set of headphones that are almost unusable because they boost the highs and crispness of the performance. It’s painful. I would not mix anything on those.

Koz

You have good ears. My computer, like most computers, has multiple fans. Intake, outflow, a fan on the processor inside and a small fan on the graphics card. It all blends together for me, I can’t single them out. Noise reduction at 13 decibels works wonders for me and I don’t hear the fan noise on my published videos. i have tried placing objects like blankets between the computer tower and my monitor/mic, but that was in place when I did the recording. I could potentially get a quieter computer in the near future, but it probably won’t work miracles. Recording video and gaming and recording on a mic taxes processors, and by extension fans. Plus I know you mentioned extension cords, but I really can’t feasibly move the computer away. The space is not 100% devoted to my recording.

I can turn up the gain on the mic, but the issue I have is that soft noises, like my breathing and mouth clicks, as well as like you mentioned airplanes outside, dogs barking, etc. are amplified along with everything else, whereas with the low gain I can speak louder, but nothing else but my voice increases in volume. I could easily make the sound 3x louder. I have both my mic’s physical gain and windows’ (Audacity uses this) microphone sensitivity way down. Is there any problem with recording “low” and just using a normalizer/leveler? On a technical level, with the output level the same, what changes are made to sound quality by doing this?

I have a pair of Audio Technica studio headphones, and like you say, listening to my work on that makes me want to pull my hair out, but I tested it and so does pretty much every gaming commentator, even those who have millions of views and literally make their living game commentating. So I never use those, but rather my computer speakers with subwoofer. I always had a nagging thought these speakers were too treble-y, but they sound find when say playing a game or listening to a song. I am curious though, my headphones amplify the echo almost unbearably. Did you not hear any echo in the sample? As I mentioned in the previous post, by the room set up behind, I don’t know what to fix. The mic shield protects the back of it, so if the echo is coming from behind me, then that wouldn’t even help.

You think the overall richness of the sound is ok? It doesn’t sound flat to you? Did the youtube samples I posted?

So you agree with my policy of lowering the treble by a bit? I do 3khz, but if you think 4 is better I will do that.

Thanks for your help! I appreciate you taking the time. I suppose the best way to boost my audio quality from what you say is to find a way to keep the sound from getting from my computer 3 ft to the right to my mic, or hoping a quieter computer is substantially quieter.

-Marcus

There’s no echo. There’s motor noise, white noise, but no echo. So it must be your computer playing it twice, with a slight delay between both signals.

Could this motor noise be vibration coupled? Have you tried putting the mic on a heavy book with a thick towel under it?

There’s no echo.

And there in a short sentence is one important problem. We’re looking at each other saying, “What’s he talking about? I don’t hear any of those problems.”

Whatever you’re doing for monitoring is lying to you.

And yes, this is one shortcoming of doing troubleshooting across multiple timezones. Now that you posted a simple sound clip, it’s obvious that you’re not listening to the same thing we are and that difference ripples through the whole forum thread. Remember up there when I said I liked different sample clips than you did and I thought you had the descriptions backwards?

I’m not sure we can go any further. You can take our recommendations on blind faith, but I wouldn’t. The object is for you to generate your own corrections so you’re happy with your own production. That would be rough to do with monitors that don’t work right.

Koz

Hi Cyrano, the mic is on a boom with a shock mount so I don’t think mic vibration is an issue.

Kozikowski, when you say monitors, what do you mean? My headphones? Or my computer in general? One thing I always do is I either record in mono or if I’m forced to record in stereo I have audacity compress to mono.

I’m interested in your advice because a) you’re experts in this field with a lot more experience and knowledge (I could fool around with the audacity settings for months and still not know what it all means) and b) you’re not me, and it’s always good to get a second opinion.

My computer is old. My motherboard (and it’s onboard soundcard and USB system) is old, so I’m actually curious to see if I do get a new computer with the newest hardware if the audio is improved at all. Again though, my headphones and speakers play out game audio just fine, so funny how it gives my voice only the echo and weird flat sound.

I’m speaking about 6 inches from the mic and sometimes I feel that’s too close, but if I move back (or if I’m focusing on the screen and so my mouth isn’t facing the mic directly) the echo is more pronounced. The main reason I was considering the shield, which is large and pricey, was because I thought it would let me move the microphone a bit more away from me and that would sound more natural. When you hear top LPers like say this guy Arumba, he sounds like he’s just in the room talking to you, not like he’s eating the mic, and I sometimes feel that you can hear that I’m right on the mic. But I’m curious about how mechanically a shield that blocs the rear and sides of the mic will help my sound when you note the echo is coming from behind me.

Thanks!!

-Marcus

funny how it gives my voice only the echo and weird flat sound.

And you need to resolve that. We don’t hear the weird, flat echo sound, so you’re producing a show you can’t hear.

So I stand corrected, the machine shop in the background is not job one. This is. The machine shop is job two.

Koz

Hey Kozikowski, would you mind if I email you some follow up questions, along with a photo of my recording area for your input?

-M

It’s generally far more valuable to keep things on the forum. The other elves are available to jump in if there is more information on any point or a correction. Plus, the posts are available to help others.

We know what the two important problems are: You’re recording in a machine shop and your monitors aren’t accurate. I have no magic past what appears here.

Koz

Ok. It was more about me not feeling comfortable posting photos of the inside of my house to the entire world, which I still won’t do.

Nevertheless I have very much appreciated your time and insight. Thank you!

-M