Audio editing order of operations (i.e. compressor, levels, normalize, etc)

[budha]

Hello everyone.

Is there a specific step by step process you should do when post processing narration/voice over audio as good as possible? I am recording podcasts for work now and I want them to sound great.

I understand you need a good mic, etc, but I am wondering what you do once it's recorded (I already have a good USB mic). I'm using STP3.

Also, is there a good way to make my voice sound richer/deeper, with an effect?

So for example (this probably isn't right, just throwing a list out there):

1. remove bg noise
2. adjust audio levels
3. run a limiter effect

Thoughts? Thanks!

 

[Noiz2]

1. Don't record in a noisy place
2. avoid noise reduction.

EQ might help with your voice, you could pitch shift it if you really want to sound big and deep, but you won't sound like you anymore.

Levels yes, normalize NO.

VO's are generally more compressed than regular dialog to make them cut through better.
You can slap a limiter on if your worried about going over but the compressor is probably a better tool. Basically the same thing but with more flexibility.

Since this is a podcast I would probably compress even more than normal and maybe use some sort of maximiser to get as much signal as poss going into the compression stage so that what comes out is still decent. A lot of podcasts get this Short Wave radio kind of sound from having levels too low and getting a ton of compression artifacts.

 

[AwakenedFilms]

Generally, start with clean audio. A 'USB' mic is a good start, but not the best option for high quality VOs. If you are a man (which can be assumed by your desires to sound deeper), you may be better off with a dynamic mic such as an RE20 through a high quality preamp and A/D interface. The closer you get to a dynamic (except for omnis) the more bass will be introduced - called proximity effect.

EQ can also add bass, but it is always more advisable to remove the frequencies you don't want rather than add the frequencies you do want.

As Noiz2 states, compression and possibly even some limiting is good for VOs but can dramatically increase the level of BG noise if over used (and add harmonic distortion).

As far as order, it is important to some degree. Generally I would start with an
1) Expander (only if absolutely necessary) its better to retake than use this
2) High Pass - set to around 75hz, closer to 100hz on females
3) Low Pass - set to around 15khz, higher if hiss is not a problem, lower if you want to clear out the problematic hiss
4) Compressor - set to mild compression such as 2.5:1, threshold of -12 or so, adding gain to preserve the audio level
5) Limiter (only if wanting real punch) - set to as much as 20:1, threshold of -4 or so.


You can also parametric EQ out certain frequencies that are problematic. Set the Q low (1), set the gain high(+12) and sweep the frequency until you hear the problems accentuated. Then reset the Q to the desired spread and bring the gain into negative territory to taste.

Most important lesson is to get clean audio to begin with - good mic in a bad environment will still sound bad. Gain staging is important. Eliminate computer noise. Minimize poor acoustics of a room.


Jason

 

[budha]

I really appreciate the help guys!

Is there an effect that would stop distortion from happening in STP? I know that FCP has the "mark" function that hits all of the peaks, but I was just wondering if STP had something that stopped anything over 0 db from clipping.

Thanks.

 

[Alex H.]

Is there an effect that would stop distortion from happening in STP? I know that FCP has the "mark" function that hits all of the peaks, but I was just wondering if STP had something that stopped anything over 0 db from clipping.

This is the digital audio world. There is no such thing as "above" 0dB, as 0dB in Full Scale means that every single bit in every single word is occupied and there is no more room for audio data. Anything that tries to cross that line is going to distort.

It's important that you set your record levels and keep both eyes and ears on the signal as you record. Give yourself plenty of headroom... record around -12dB with occasional peaks at -6dB.

 

[budha]

C2V,

Sorry I wasn't clear enough. I didn't record the audio, someone else did, and there is an occasional peak that hits red every once in a while. I am curious if there is a simple way to fix this in STP.

 

[Alex H.]

Well, using your mix levels will allow you to lower the level for peaks... compression and limiting will also help tame peaks. As for distortion, if it was recorded peaking at 0dB it's likely that the distortion is there to stay even after levels have been mixed.

 

[Noiz2]

The only thing I have found at all effective is Rx by iZotope. But that is no guarantee and you still don't get anything back, but it can often smoothe out the clipped parts and make them sound more analog clipped rather that digital clipped and that is a big improvement.

 

[Tommy K]

Hello everyone.

Is there a specific step by step process you should do when post processing narration/voice over audio as good as possible? I am recording podcasts for work now and I want them to sound great. Thoughts? Thanks!

Here is a step procedure that works very well.

1. Normalize audio to around -13 dB, it depends on what levels were captured, you may need more or less... Use "RMS" normalization... not peak. Most normalization filters in ediitng apps effect only peak. RMS normalization (Root Mean Square) will raise the lower portions up and even the sound out more... meaning it averages it and boosts the lower levels. Peak will not do this... If your normalize filter has it... select "dynamic compression if clipping occurs" to avoid clipping the signal. This will make sure your peaks will not go over 0 dB and distort during normalization.

2. Equalize... Cut the frequencies below 200 Hz to infinity. Do the same thing for above 15 Khz. Boost the 2.5 Khz to 3.5 Khz range by 3dB. This will add a little punch to the voice.

3. Noise reduction.. if needed use it sparingly at around -12 dB reduction. Select a silent part of the voice clip and use that for your noise print. This will help with any extraneous hiss, hum etc hidden in the voice track and during silence...

4. Compression/limiting... apply a mild compresssion level of about 1.5 to 1 and set your limiter to what ever the max audio level you need. I always set my limiter to -2 or -3 dB for final output of all my audio. Lots of level and no clipping or distortion. This way I know exactly what all of my levels are throughout the entire program.

This recipe will definitely enhance your sound and keep it even throughout. Do not over normalize.

 

[budha]

Thanks for the help everyone!

 

[Noiz2]

Here is a step procedure that works very well.

1. Normalize audio to around -13 dB, it depends on what levels were captured, you may need more or less... Use "RMS" normalization... not peak. Most normalization filters in ediitng apps effect only peak. RMS normalization (Root Mean Square) will raise the lower portions up and even the sound out more... meaning it averages it and boosts the lower levels. Peak will not do this... If your normalize filter has it... select "dynamic compression if clipping occurs" to avoid clipping the signal. This will make sure your peaks will not go over 0 dB and distort during normalization.

EEEEK. Not quite correct. All Normalizers raise all levels. Peak VS RMS is just what your basing the normalization on, the average level or the peak level. Personally I rarely use them because they have a tendency to make your tracks flat. People talk at different levels you don't want them all the same.

2. Equalize... Cut the frequencies below 200 Hz to infinity. Do the same thing for above 15 Khz. Boost the 2.5 Khz to 3.5 Khz range by 3dB. This will add a little punch to the voice.

EEEK again. Well not a big one but depending on the voice 200Hz might be a bit high. Also depending on the tracks that might be a bit much of a boost.

3. Noise reduction.. if needed use it sparingly at around -12 dB reduction. Select a silent part of the voice clip and use that for your noise print. This will help with any extraneous hiss, hum etc hidden in the voice track and during silence...

Unless your using Rx that is a lot of NR. If your using sat Digi's BNR -12 dB is WAY too much.

4. Compression/limiting... apply a mild compresssion level of about 1.5 to 1 and set your limiter to what ever the max audio level you need. I always set my limiter to -2 or -3 dB for final output of all my audio. Lots of level and no clipping or distortion. This way I know exactly what all of my levels are throughout the entire program.

That's a bit high. Most places don't want to see peaks over -6 and anything over -3 might clip because of D/A over shoot. I generally would have it set around -6 or for broadcast -12. Also you want to make sure your not hitting it too hard.

This recipe will definitely enhance your sound and keep it even throughout. Do not over normalize.

We may be using somewhat different definitions but you don't want your sound "even throughout". Dynamic range is your friend, it keeps the ears from getting bored.

 

[Tommy K]

I got this recipe for voice from a very good sound engineer a while back. I have used this or a slight variation for many broadcast TV programs with very good results... Never had a complaint It just depends on what the audio levels are too begin with...

1. EEK When I normalize, I intend to raise every level that I intend to raise... RMS is the choice. I can adjust what level gets raised and where my floor is at. If I don't need to boost the levels I won't add it.

2. EEk -12 dB NR is certainly not to much noise reduction... This works great in Sound Forge and other apps.. You may need less you may need more... depending on what you are trying to get rid of... and how much separation you need.

3. I edit and archive all my projects audio at a limited -3 dB. I lower it when delivering a master to whatever the spec they require.

 

[Noiz2]

1. EEK When I normalize, I intend to raise every level that I intend to raise... RMS is the choice. I can adjust what level gets raised and where my floor is at. If I don't need to boost the levels I won't add it.

?? "I intend to raise every level that I intend to raise..." If you normalize you raise ALL levels. The only difference with using a straight level change is that the plugin figures out how much the level needs to be raised or lowered to hit your "goal" level. The danger with using an RMS "goal" is that you could end up clipping since it's boosting the over all level to make the average at X.

That said it is possible that Sound Forge is not actually "normalizing" but doing some kind of "maximizing" and so is limiting the max level. Companies sometimes use technically incorrect names for things so?
This "normalize" function sounds a lot like a "maximizer" rather than a "normalizer".


It should also probably be noted that how much of a good thing or a bad thing "normalizing" is on your tracks has a lot to do with exactly what your trying to do. The OP asked about dialog editing and I was thinking along feature film lines. TV is generally much more compressed that films and VO and that sort of thing are generally VERY compressed. In those cases yes you want to "maximize" the hell out of the track. But generally on feature films there is very gentle compression and "normalizing" is not at all the norm.

2. EEk -12 dB NR is certainly not to much noise reduction... This works great in Sound Forge and other apps.. You may need less you may need more... depending on what you are trying to get rid of... and how much separation you need.

Again this depends a bit. But -12 is a LOT of noise reduction. If what your taking out is very light then you can get away with more NR since your noise print is not very intrusive on what you want to keep. It also depends a LOT on who's NR software you are using. Sound Forge has I believe (it's been many years since I used it) a very good NR function. Rx can often work well at -12, but that is the only NR app that I have used that can go -12 with out adding a lot of artifacts. So you mat regularly get away with -12 but as a general note for people new to this I would be leery of going that strong as a "rule".

Not sure what you mean by "separation" and what that would have to do with NR.

3. I edit and archive all my projects audio at a limited -3 dB. I lower it when delivering a master to whatever the spec they require.

SF is a two track editor, or was anyway, so you would be making destructive masters. Post for film is done with multi track DAW's so your archive is the session/project not the output file so I'm only outputting the master. So this is mostly a difference in work flow. If I was archiving a stereo master then I might do the same thing, though I would probably archive at whatever my working level was. If I were archiving on analog tape I would push the levels but as a digital file there isn't much advantage in pushing the level and there is the (slight now days) chance of degradation because of the extra processing. I suspect the last is an "old school" worry from when the math in volume changes wasn't as high quality as it is today and you could pick up math errors that were audible. That is also one of the reasons that "normalizing" got such a bad rep with dialog editors.

 

[Tommy K]

Noiz2... I have to be honest here... your post with all the EEkS! was a bit confrontational right from the start. Was this necessary in order to give your opinion?

OK... Why are you commenting on Sound Forge if you don't use it or know it's true functions? I use Sound Forge and have used it since the Sonic Foundry days... back in the 90's. The person who gave me this recipe for "VOICE" was on the Sound Forge developmental team at Sonic Foundry (before Sony bought them out in 2003). This was his recipe for primarily low or uneven VOICES in interview situations or where someon didn't do a good job micing a person etc. It is not intended for feature film or situations where high dynamic range and low/no compression is desireable. If you need more punch, this will do it. This recipe is for unifying levels and compressing different voice levels, as well as getting rid of unwanted low freq rumbles and/or high freq hiss, ambient noises etc. It works and it works well WITH OUT artifacts using Sound Forge...

What did I mean by separation? Putting some range between the voice and the audible level of undesired noise. Such as... if the voice level is set to -6 dB.. I would like the noise to be at least 12 dB down or greater from the voice. Akin to signal to noise ration as an analogy... I don't want to hear it... Clear?

Sound Forge is non destructive... You simply "SAVE AS". Sound Forge is indeed true normalization with either PEAK or RMS choices.. with built in dynamic compression to avoid clipping if desired and selected. The maximizer plugin in Sound Forge is called the "WAVEHAMMER"... which is for limiting max levels.

The NR filter is a very good one with different NR modes. Yep in the wrong hands you can easily produce lots of unwanted artifacts.

I am not an audio engineer.. But I have worked in electrical engineering for 20 years including the space shuttle program... So you aren't schooling a newb here... I understand the video and audio technical issues of film making quite well... My post was meant to help someone who could use this proven worlflow for it's intended purpose... not to get into a long winded debate.

Have a great day

 

[JordanBlock]

Noiz2... I have to be honest here... your post with all the EEkS! was a bit confrontational right from the start. Was this necessary in order to give your opinion?

FWIW, My response to using the posted recipe as a starting point was 'eek' too.

Sound Forge is non destructive... You simply "SAVE AS".

Sorry, it's a destructive editor. You of course have the option of saving to a different file if you choose, but it's still by definition a destructive editor.

 

[Tommy K]

FWIW, My response to using the posted recipe as a starting point was 'eek' too.

Sorry, it's a destructive editor. You of course have the option of saving to a different file if you choose, but it's still by definition a destructive editor.

Gotta love the internet... I know many people who use this recipe and are very happy with the results... I just used this technique today on some interview clips I received from a client that had uneven and low audio. Hmmm seemed to work out quite well actually... They really liked the improvements.

EEK... Maybe you need to learn more about when, and when not to use it?

 

[JordanBlock]

Gotta love the internet... I know many people who use this recipe and are very happy with the results... I just used this technique today on some interview clips I received from a client that had uneven and low audio. Hmmm seemed to work out quite well actually... They really liked the improvements.

EEK... Maybe you need to learn more about when, and when not to use it?

If it's working for you, great. The settings posted would probably do an ok job making poorly recorded audio more passable.

With properly recorded audio, they'd probably be doing more harm than good.

The correct answer in so many situations with sound is 'it depends'. There's just no one magic formula of filters.

 

[Tommy K]

If it's working for you, great. The settings posted would probably do an ok job making poorly recorded audio more passable.

With properly recorded audio, they'd probably be doing more harm than good.

The correct answer in so many situations with sound is 'it depends'. There's just no one magic formula of filters.

It depends is a good answer... We agree... I believe I stated a couple of times that this was helpful for low or uneven levels. If you already have beefy audio levels than you probably won't want to normalize. But you can still EQ and do NR if you have any offending background noise.. and you can limit the output to a safe uniform peak level which is always a good idea....

The OP asked about which order and the way to compress, normalize etc. I gave him a proven work flow step by step... You need to decide if this will enhance your audio or harm it...

 

[Noiz2]

Noiz2... I have to be honest here... your post with all the EEkS! was a bit confrontational right from the start. Was this necessary in order to give your opinion?

OK, I think your suggestions are a bad idea 99% of the time. Is that better? No well maybe it wasn't nec. I thought it was less confrontational than going off on some of the suggestions that I think in general are just terrible suggestions.

To be fair though this is a film/video sound forum and the guy did ask about podcasts so for a podcast then fine do what you want. It's going to be compressed to hell and gone anyway.

People responded, I think, appropriately for film/video. And that is NO you don't want to use aggressive noise reduction, NO you don't want to use a normalizer, NO you don't want to have your tracks at -3, etc.

OK... Why are you commenting on Sound Forge if you don't use it or know it's true functions? I use Sound Forge and have used it since the Sonic Foundry days... back in the 90's. The person who gave me this recipe for "VOICE" was on the Sound Forge developmental team at Sonic Foundry (before Sony bought them out in 2003).

Now who's not listening? I did use Sound Forge back then. And I had all the add ons also. However sound for post is done 99% in Pro Tools and at that time Pro Tools didn't run on the PC so I had to switch to a Mac and to Pro Tools. But even then I didn't use a destructive two track editor for post. I used SAW, a really nice DAW for Windows. Post is not done on a destructive editing system by anyone who does it seriously, don't tell your clients. I wouldn't do a podcast on a two track editor and certainly not on a destructive editing system. And sorry but a software developer is probably the last person you should be taking sound advice from. That's like going to Snap-On to get your car fixed. But even assuming they were one of the sound guy testers, they were testing a two track destructive editor and the tool for what the OP wants to do is a DAW not a sound file editor.

This was his recipe for primarily low or uneven VOICES in interview situations or where someon didn't do a good job micing a person etc. It is not intended for feature film or situations where high dynamic range and low/no compression is desireable.

That would be a valid point if the OP had asked for ways to punch up "low or uneven VOICES", but he didn't.

If you need more punch, this will do it. This recipe is for unifying levels and compressing different voice levels, as well as getting rid of unwanted low freq rumbles and/or high freq hiss, ambient noises etc. It works and it works well WITH OUT artifacts using Sound Forge...

I will believe that when I hear it. I really doubt that I wouldn't hear the artifacts, but?

What did I mean by separation? Putting some range between the voice and the audible level of undesired noise. Such as... if the voice level is set to -6 dB.. I would like the noise to be at least 12 dB down or greater from the voice. Akin to signal to noise ration as an analogy... I don't want to hear it... Clear?

If your noise is at -12 you have some serious problems. In fact if you have gotten this by lowering your noise by -12, your nose was four times as loud as your dialog to start out with????? But I get what your talking about, I have just never heard it described as "separation".

Sound Forge is non destructive... You simply "SAVE AS". Sound Forge is indeed true normalization with either PEAK or RMS choices.. with built in dynamic compression to avoid clipping if desired and selected. The maximizer plugin in Sound Forge is called the "WAVEHAMMER"... which is for limiting max levels.

WRONG That is not what nondestructive means.
If SF is doing any dynamics processing then it is not a normalize function as everyone else defines it.
A maximizer is not a Limiter, or rather it is more than a limiter. With out a current copy of SF to look at I can't tell if your misunderstanding what these tools are actually doing or if SF is using non standard terminology. Limiter , Maximizer, and Normalize have standard definitions that the vast majority of applications adhere to. It wouldn't shock me if Sony decided to "think different" and redefine them but you should know them. If your trying to show how knowledgeable you are then why are you using the terms incorrectly?

I am not an audio engineer.. But I have worked in electrical engineering for 20 years including the space shuttle program... So you aren't schooling a newb here... I understand the video and audio technical issues of film making quite well... My post was meant to help someone who could use this proven worlflow for it's intended purpose... not to get into a long winded debate.

So that gives you credentials in sound post how? I have worked in sound about 35 years now. Theatre, Music, Video and big and low budget films. I am not an electrical engineer but I did build the first two mixers I owned, and the first computer I owned. And my brother in-law worked at mission control, but that has nothing to do with the OP question. From your posts I would say that as far as post audio goes you are pretty close to a newb, and if we were talking electrical engineering I would quickly find out how close to a newb I was (very close). Theory is not practice and often has only a tentative connection. If you really were in the thick of it you probably wouldn't be saying things like "I understand the video and audio technical issues of film making quite well", because it's a moving target and the "issues" today were mostly not even on the radar ten years ago, and most not five years ago. I deal with it every day and spend a LOT of time keeping informed and I probably wouldn't say "I understand the video and audio technical issues of film making quite well", because between when I started this post and am now typing some new codec just came out and will make my life hell tomorrow. So if you actually know how all the new codecs and new workflows mix with the audio and video and film go together in all their multitudes of variations, my hats off to you and I grovel at the alter of your greatness.

Bottom line. If you have something that is working for you then go for it. Go ahead and share it. But in the end that is all it is, a formula that is working for you in your particular niche. I've put my opinions out there also and they are the same thing. Stuff that has and/or has not worked for me in the specific situations I have been in. Each situation is a bit different. My experience makes me very skeptical of your formula, and my income depends on my judgment so I. going to probably shy away from it. But that doesn't men it is actually wrong in all cases, or that I think it's wrong in all cases. But based on my experience it's not something I would likely do. I wasn't trying to offend, or confront and I did give my reasons why I had problems with some of your suggestions.

Have a great day

I'm taking that literally and returning it.

Cheers
SK

 

[Tommy K]

Noize... wow ... that is all I can say to that dissertation above lol. I'm sorry you are having such a bad day. Sit down take a deep breath lol. I'm sure you are a great guy... You took some fairly harsh shots at me... not real classy on your part... I won't be drawn into that type of behavior... enough said. I come here to LEARN and help out when I can. PERIOD.

Yes I was wrong in not saying that SF is destructive. BUT... I was actually being sarcastic and saying use the "save as"... Than you won't destroy the original file... I guess it didn't come across that way. I'm well aware of the difference. It's a small annoyance, if any to me. SF is a great little app.

No need to "grovel at my alter of greatness". Great I am not. Humility seems to be a lost virtue these days. I am particularly fond of it myself. I have to learn new things constantly like everyone else. I try to study almost everyday to try and keep up with film making and general technology. I AM NOT A SOUND ENGINEER and never have been. I made that clear. I was NTSC TV and Radio Frequency/Microwave specialist. My only point in stating that I have an EE background was to explain that I understand most of the technical jargon bandied about... and not starting from square one. That was my sole purpose my friend.


I am now posting word for word the recipe for low or on camera mike audio that was sent to me from a member of the Sound Forge development team. It is not my recipe... But it has been useful many times to me and others... If this applies to your particular situation, it works well for what it is intended...


From: JimF@SonicFoundry.com

Hey Tom, here is a great way to clean up DV audio. If you have video that was recorded with low audio or using an on camera mic at an event, it's going to need some massaging to sound good. Import the audio track into Sound Forge and get to work! I usually start with :

1. Process> Equalizer> Paragraphic, check both the "Enable Low Shelf at 100 Hz at -inf" and "High Shelf at 10,000 at -inf". This gets rid of unnecessary low and high end noise.

Note.... If there is a troublesome frequency band, say from some machinery, I'll find it with this same EQ by boosting and sweeping, than cutting once I hear the offending sound.

Note: depending on the level of your recorded audio you may or may not have to use the normalize function to raise it to a useful level. If you do, tweak the amount so the file is raised but does not start clipping. Select the "Use Dynamic Compression" if clipping occurs box.

2. If you have it, use the Noise Reduction plug in to cut back on extraneous noise.

Note: If you would have some bad background noise, you might be able to follow up that treatment with "Effects> Noise Gate and smoothly eliminate the background (except when the voice is speaking).


3. Go back to the Paragraphic EQ and put a little 2 to 4 dB bump between 2.5 KHz and 3.5 KHz to make speech more intelligible.

4. Finally even out the levels with some dynamic processing. Go to Effects> Wave Hammer and select "Master for 16 bit". You can tweak this if you like to make your file louder or lower, and make it even. The default is around 1.5 to 1 compression and is mild. You can limit the level of the file using the "Volume Maximizer" I usually set it so nothing peaks above -3 dB to keep things on the safe side.

Happy Forging!
Jim B