I've been handed a butt load of narration.

It was all recorded on a microphone without anything but the microphone.

And the narrator didn't move the head closer or farther away at any point.

So I'm sitting here with levels that for the most part are average peaking at -9 and dropping way below -20db.

I'm trying like hell to come up with a list of things I can do to bring the recordings up without having distortion creep it's way in.

I thought I had it with a 4:1 compression and work the output for each individual phrase, then a slight EQ but dang if I'm finding a burn hiss creep in from time to time.

What can I do? I've told the director today that the narrator stuff has to be rerecorded and that I'll spend time with that person to make sure they understand proper recording value's, get them recording stuff thats peaking around -3, not -9 to -12.

Am I on the right track? Or is there something I just don't know that can fix this situation?

I don't hear any background hiss on the recordings but there's almost no dynamics to the raw files, really high end, sharp and needing to have the high frequencies cut substantially.

I can normalize without getting any hiss...should I compress then?

Signed...been trying to fix this for a week...ugh.

First thing I would try to do, if I had to use the recording, is to edit out all of the spaces between words to digital silence. This can be done automatically or by hand depending on the part. If the levels are too soft in places then it would be best to move the problem words or phrases to a separate track and use a gain plugin to bring them up to a better level for 'strip silence' ( as it is called in logic, but most DAWs will have a similar function) to do it's thing.
Once you have cleaned the dialogue up you can try compression, maybe using more than one in series, or an automatic leveller such as waves 'vocal rider". Eq may be placed before and/or after, depending on source problems, e.g. Proximity effect can be filtered out before any processing.

Yep first thing I did was silenced out...that's habit for me.

By gain plug in do you mean basically normalizing? Bringing it up to 0db? -1?

I'm old school with my editor, (cool edit).

What I'm basically finding with any different combination of tries, is that there seems to be almost a background brushing sound. I know this is coming from pulling the vocal up beyond where it was recorded at and my gut is telling me that I'm the one creating this distortion effect....

....almost as if I'm stretching the balloon past it's manufacturer specs if you will.

Am I on the right track thinking this? And the reason I ask is from the stand point of hours to be wasted trying to find solutions, only finding that same odd background subtle brush sound.

What distortion is creeping in - quantisation distortion? is it just a matter of it being recorded far too quietly, or is the guy too far from the mic?

Whenever I've had to work with bad dialogue recordings, low level isn't usually a problem - the problems usually come from poor equipment (ie lots of hiss), too far from mic, or unwanted room tone.

Failing to compress or recording at too low a volume shouldn't be an issue (unless it's stupidly low, which -20 isn't usually - after all, it's around 0VU with most calibrations!).

Strip silence can help but if it's really noisy, you'll just notice the noise coming in under the vocal.

Recorded far too quietly. And probably too far from the microphone as well.

I believe it is cheap microphone and no other piece of equipment and recorded into something called Audacity I believe.

I'm rerecording them into my cool edit and with everything I've got to jack the signal I can barely get -15-20 in some of the later recordings.

I don't know what quantisation distortion means but I can tell you that the sound I get is only after I try bringing up the levels by normalizing, or by compression..I'm trying all sorts of different compression values...I'm trying all sorts of different EQ settings, parametric etc...

But after I get everything to maybe I'm happy and listening back in odd few words or phrases I hear a very subtle Ffffft (like white noise on a tv channel...but really really hardly there and only briefly and only in a the odd word or so).

There's no background noise to begin with, there's no hiss problem.

Just that odd sound and again it makes me think it's a limitation of the equipment based on the low recorded levels and me jacking everything up.

I guess the main question I have is what is the methodology at this point?

Do I get a rerecording done properly? Or do I keep at it, keep trying different combinations to see if I can tease up those levels without getting that brushing sound?

Audacity is just software, it's not a problem.

When you say "re-recording" into Cool edit, what do you mean? why not just start off by importing the WAV files (is that what they've given you?) and gaining them up so you've got a bit of level to work with? You can ALWAYS get the signal as loud as you want - if it sounds good or not is another thing.

I wouldn't be trying to add gain in the analogue domain or anything like that.

But if it's a cheap mic (and presumably cheap preamp - must have been SOME other equipment unless it's a USB mic) and recorded with room tone, there's not a great deal you can do.

The noise you're hearing...might be quantisation distortion, might be some noise in the room (or even in his voice) that's only audible once you've jacked it up to a decent level.

If you are just normalizing then any distortion you are hearing will be part of the recording, and you are probably stuck with it.
It is difficult to repair a distorted recording enough to make it anything like a good clean one. Eqing or deessing can only do so much before it begins to sound unnatural.
Try normalising first and listen for any problems which are part of the recording, then plan from there. If the recording is ok, then it is best to work with the audio at around -12db to allow enough headroom for any further processing. If you hear distortion, then it is best to deal with it first, or any compression will just make it more obvious. You might have to filter out some frequencies, or use a dynamic filter or eq, like a deesser.

Couple of questions:
Firstly, what format are these recordings in? If they are digital files like WAV or AIFF then it is best to import them directly into your editor, unless they are at such a low level that they need to be amplified.
Secondly, if they are digital, what is the sample rate and bit rate? This will need to be taken into account before you import them.

If they've already been digitised (which of course they have been), however quiet they are, it's always best to add gain in the digital domain...you've got nothing to gain (pun intended) by taking them out through conversion (at a low level) and using an amplifier to boost them.

Yes, you are right. Keeping it in the digital domain is better.

No unfortunately I have mp3's of each take. So I use an mp3 player to play back as I record in my Cool Edit to work on cleaning and what not on each take.

I will be speaking with the narrator next week, I don't think the equipment is top notch, something tells me it's a bottom line Shurr with a USB but that's a guess.

I've stumbled across a combination that might work. First I'm normalizing each word or phrase to 0db, run a hard desser compression, run a parametric highs softener with my output to as close to 0db as possible...

Then a realaudio compander compression again with the output as close to 0db as possible, hard limit everything to -6db and then finally normalize to -3db.

Seems to sound fine with a few minor glitches but I'm working on one of the quietest of the recorded segments.

I appreciate the help thank you by the way.

Ugh they're mp3 files. I have to use an mp3 player to play them while recording them into my Cool Edit. double ugh.

I have to head to bed, it's quarter to 3am here and I'm bagged.

I mentioned above that I may have come across a combination that seems to work but I wanted to say thanks for trying to help, appreciated!!

Why not convert them to wav and import into Cool Edit?

edit: I did not see that about Audacity and I thought they'd been recorded directly to MP3 through some portable device. If there is any way for you to get hold of the original wavs, use the original wavs! It will save you a lot of fuss.

Here is the technique that I use:
1. Auto-level the dialog with Waves Vocal Rider (or HoRNet AutoGain)
2. Apply noise reduction (Waves NS1 / W43 / WNS / Z-Noise or iZotope RX 2)
3. Apply corrective EQ, if necessary
4. Apply compression, if necessary

To remove room noise, I use iZotope RX 2 in standalone mode.

To remove reverberation, I use Zynaptiq UNVEIL.

To *add* room noise for ADR matching, I use Zynaptiq UNFILTER.

Can't Cool Edit import MP3s? If not, use a digital (software) converter and then import. Playing the files from an MP3 player and recording introduces a whole raft of other (possible) problems.

r,
j,

Yeah I only got that odd sound after I did a typical EQ band with some bottom end. Then it just seemed to appear...wasn't there until then which means it was always there but highlighting the bottom end just brought it out.

I went about it differently, attacked it from taking down the highs and adding very little to lows...seemed to work

Well, there's a big part of the problem. Cheap converters in, cheap converters out (plus a probable cheap headphone amp), conversion in again.

If they recorded into audacity, at some point it will have been a WAV file. You're wasting your time on the MP3s + player - stop working on this immediately, there must be a better source available. It's typical "export as MP3 because the guy self-recording doesn't know any better". But he must be able to at least provide you the WAV.

Still should be able to get a useable recording.

I just would stop work right now. You're wasting your time. At least get the WAVs first (and if for some reason he's deleted them, get him to re-record it - if your results are then re-MP3ed, it'll sound even worse).


Glad it's helping - maybe a bit of moral support at least! I feel your pain....

Nope Cool doesn't import mp3 but I like your suggestion and I'll try that tomorrow thanks!

Sounds like he's been given a player loaded with the MP3s....see my more detailed reply above!

Ok cool suggestions, I'll talk to the director tomorrow about the original wav files. thanks again!

This looks like an awesome audio recipe, thanks! I'll keep this and play around with this at some point. Appreciate the sharing.

Oh dear. As a last resort, perhaps it can be connected as a USB device, so that the files can be transferred digitally.

OP: what kind of MP3 device did you get?

r,
j,

To be able to provide something that will translate back to MP3 again (which also means MPEGs, Youtube or DVD - audio on a DVD is just a good MP3) you need to - otherwise you're going to trash the audio as soon as it's converted back. Regardless of how good it sounds as a WAV.

Even so...working from MP3 for narration (presumably for video) is not a good idea!

Oh, agreed, totally

j,

During recording, your peaks don't have to come up past -12dBfs - provided you're at 24bit or higher.

It is during mixing and mastering that you want to get your levels up so that you're peaking above -4.

And everyone on here is right - unless you're miking your own aunt's recipe for rhubarb pie, nothing production-wise should be done in mp3.

The root of all your troubles in this case is poor mic technique. Your client's narrator is all over the mic pickup zone, in and out!

I had to do some forensic work on some voices that had been recorded on an Olympus hand held recorder. The person chose the lowest possible sample rate and the level never got above -20 dBFS. There was so much distortion and digital artifacts I really could do nothing with the tracks. I am wondering if someone did this on a small hand held device and then transferred that to audacity and that is why you are having the problems you are having?

On a side note I just got asked by a book publisher if I could clean up some tracks that one of their authors had recorded at home to save money and I found the same problems. If people want professional results they should go to professionals IMHO. but that seem counter to what a lot of people do today. They want to go the cheap route and then find someone else to "fix" all their mistakes. It seem counterproductive but it seems like what is happening more and more today.

My "worst" story: I was tasked with cleaning up audio recorded with the built-in mic of a videocamera, including air-condition noise, audience sneezing, camera handling etc. I though I did a pretty good job with the first half hour, but the client backed out in the end. I'm actually happy, as it was painstaking work. Would have been expensive for them, though.

r,
j,

What is the final product? TV? Video? Audio Book? Film? Are you passing this off to some else to mix? Each of these scenarios are treated a little different.

What does this mean? No pre? Or just a mic and a Pre? If so, so what? I have been recording VO's like that for 15 years.

Sounds perfect! Seriously. Those are great levels. If the system is calibrated, you are not pushing the analog stage and hitting the converter cleanly. At 24 bit, heck even 16 bit, those are lovely levels.

Why??

Have you tried volume? Volume riding (usually with a fader) is the single best way to get paragraphs, sentences, phrases, words, syllables, and consonants to sit just where they need to be.

Immediately going to normalization and heavy compression is like being at a picnic and asking Gallagher to cut the watermelon for everyone.

Take your reasonable levels, grab a fader, ride the levels, and use EQ and comp. to put the finishing polish and strength to it. You can also put the comp and EQ on so you mix INTO the comp.

If you read through the thread, there's a bit more to it than simple levelling...starting with an MP3 player!