Trying to find some very specific editing directions?

[Anise]

Hey all,
I used to do readings for Librivox with Audacity a few years ago, and I'm thinking about getting back into it. First, though, I need to go over some features of Audacity to be sure I remember exactly how the recordings were done. I'm trying to find a super specific set of directions on exactly how to improve the sound of the recording and end up with the Librivox's "house sound". I've looked everywhere and I can't find this extremely specific set of directions-- not on the editing page, not on the newbie recording page, etc etc. I'm not sure where I originally found them, but these were NOT general directions about how to record and edit-- these actions would be done at the very end, after recording, and they were very pinpointed and detailed. "Push this exact button, set this exact filter on Audacity to this exact number, reduce noise in this precise way, set levels in this one specific way," and so on. Does this sound familiar to anybody??? I can also use Adobe Audition if anyone has similar directions for that, but I know there were super specific directions for getting a very precise sound from Audacity... SOMEWHERE. I must have saved notes on them, but who knows where those ended up. Thanks for any and all advice!

[knotyouraveragejo]

Here is the noise reduction page

http://wiki.librivox.org/index.php?title=Noisecleaning_With_Pics#Clean_up_the_Noise:_Audacity_versions_from_2.1.0

If you have low frequency or mains hum see this page

http://wiki.librivox.org/index.php?title=Removing_Mains_Hum

Other Audacity tutorials are linked here

http://wiki.librivox.org/index.php?title=Audacity_Tutorials

If you like video instructions instead, see this thread

viewtopic.php?f=23&t=33439&start=285

[TriciaG]

Exact numbers in the directions would differ according to one's setup and environment. So maybe at some point someone gave you exact instructions, but they wouldn't have posted them for the general public, because they'd be different for everyone else.

I found one thread where you discussed your own sound issues, but it doesn't have exact values listed: viewtopic.php?f=23&t=39464

[Anise]

Thanks-- I think I'm at least getting closer to the idea... here's a quote I found from the April 2012 thread;
"
Here it is. Compressed and amplified and noise-reduced and everything. Yay.

http://upload.librivox.org/share/uploads/tests/sample_Anise.mp3"

So I think I must have already had those specific instructions from somewhere. I can ALMOST even remember how it was done. Maybe first it was amplified, and then it was noise reduced, and then it was compressed,or....? Maybe I went through the forums and found pieces of information about how to do this and then put them all together?? The big question is: Is there some specific way that people usually do this, some correct order that usually gets recommended over and over again? "Take the completed sound file you recorded and first do X, then Y, then Z...."??

[TriciaG]

There's not really a correct order. People have different opinions on what order is best.

If your recording has big spikes, sometimes doing noise reduction before compression will cause clipped spikes. (Don't ask me why; I haven't figured it out!) So in that case, doing compression first is better. But if you're nowhere near the top/bottom of the recording window, this isn't an issue.

If your recording is usually too low and so you amplify, it's better to noise clean before amplifying, so that you don't make the noise louder, too.

OK, so if it were me, I'd either go Compression/Noise Clean/Amplify or Noise Clean/Compression/Amplify, depending on if you'll clip when noise cleaning.

But like I said, there's no concrete, solid, best order to do it.

[Peter Why]

Here's what I usually do to my recordings, but you do need to work out a pattern that suits your particular recordings.

A. Compress:- Select All/Compressor with settings: Threshhold -14dB, Floor -40dB, Ratio 5:1, Attack 0.2 sec, Release 1.0 sec. Check box for makeup gain, NOT checked "compresion based on peaks" [some of these are the default values]
B. Normalise:- Check remove DC offset, check Normalise max amplitude to -1.0dB, NOT checked normalise stereo channels independently.
C. Export mp3 for amplitude check:- Save Other/Export as mp3 (to the desktop, which is where I keep the Checker program). Drag into Checker, note the overall amplitude.
D. Amplify sound file to 89dB:- Return to Audacity file. If the mp3 wasn't close to 89db, Select All, then Effect/Amplify, and amplify +/- appropriately.
E. Noise reduction:- Select a few seconds of silence in the sound file. Effect/Noise reduction/get noise profile. Then Select All. Effect/Noise Reduction Step 2: 7dB, 6.00 sensitivity, 3 frequency smoothing bands. Check by ear or by looking at the silence using the logarithmic view (from the pull-down at the left of the sound file, choose the "wavefore (dB)" view) ... and repeat the noise reduction as needed. [I usually carry out two 7dB noise reductions.]

We all have different methods! Find one that suits you.

Peter

[Elizabby]

If you want to refine your settings, as it has been a while, I'm sure you would be welcome to submit a new 1 minute test with some questions for the technical people to check and help you refine your settings as they are now?

[mahne]

Adding that audacitys noise removal tool has changed since then (better now), so even if you found numbers from back then, they wouldn't make sense with a new audacity version. And I would strongly recommend using an up-to-date audacity version.
And if anything from equipment, setup, your voice, or even the room decoration has changed, you would also be better off with a new 1-minute test.

Cheers
mahne

[Anise]

Thanks all! I will do a new one minute test at some point. Right now is probably not the time... that's what I get for going back to school again... but I would like to get involved again soon.

[MagicMan]

@Anise. Good evening. I think you may be referring to a fx chain that works for ACX Audio Book requirements. While I do not use Audacity as my DAW, I do keep up to date on it for those who may have questions. You need to add the following items set to these specifications, in the order I have them listed. The first thing you need to do is have your input signal between a -12dbs and a -18dbs to start this process. Once you make your recording, do the following and you will hit the mark, 100% of the time with no alterations needed between chapters.

1. Set the gate to a -40db to start with. You need to find the lowest db concerning speech to make sure you do not start chopping your words and adjust your gate threshold to that number, mines 2dbs.
2. Set the High Pass Filter between 80 and 100hrz with a 48db reduction. If you have a deep voice, you may need to be around a 40 to 50 hrz. If you are a female, this setting will work great.
3. Set the RMS Normalize to a -19.5.
4. Set the limiter to a -3.5dbs.
5. Run the DC Offset at the default level.

EDIT: I SWITCHED OUT THE LAST VIDEO WITH A NEW ONE AFTER DOING THE MATH. PLEASE LET ME KNOW IF IT IS OK.

That is your ACX FX Chain. Once you finish your recording and apply this chain and your last step will be to either gate out the noise floor or use the noise removal tool. I did a video for you to look at, to see what I am talking about. It also deals with eqing as well as compression. I was in a hurry so please forgive the clicks, I am coming off of a 48 hr work block and my eyelids are not cooperating.

[annise]

However - the fx chain would not conform to LibriVox technical standards - I know the bitrate is different , other things probably are too

[MagicMan]

From the information I have read in the forum, the following parameters need to be met. I apologize in advance if I have missed something as I have never used this type of audio measuring system before. The volume target is 89 but, if it is between 87 and 92 that is except-able as well. The file needs to be recorded in mono at 44100Hz and exported at 128 kbps in a mono file as well. I could not find any information concerning the true peak requirements.

If you change the RMS normalized to 20.5 from 19.5 as mentioned in my above post, this should meet the requirements mentioned in the above paragraph. I did this test run this morning and this is what the pic shows below. I am using version 2.1.3 of Audacity.