Wanted: mastering chain for meeting ACX Guidelines

Also, the waveform after normalizing looks nothing like the mastering “video resource” example they give.

What video might that be?

Ok, yay, the noise floor conforms, even with the laptop. I totally get it - there will be no removing the laptop in post, because the tone is right in the middle. It won’t filter out. For here on I’ll record without it.

At the risk of diverting you from the Bright and Shining Path (leave the laptop off) you can partially take that noise out of the show. Noise Reduction has a technique of removing noises everywhere but your voice. This isn’t perfection, but it helps. If you have a background hiss or other constant noise, having Noise Reduction remove it from between your words makes it seem like you have a quiet room when you really don’t. This prevents science fiction voices which is what would happen if Noise Reduction tried to suck medium pitch musical tones from your voice. Carried to extremes, it fails. If you have high hiss level, Noise Reduction will give you hissy snake voices. Or SciFi voices. Your pick.

I’m calling it Noise Reduction through all this because that’s what it will be in the next version of Audacity. Too many people expect Noise Removal to, y’know, Remove Noise. To nothing.

I’m pleased to look at one of your troublesome finished pieces. How? The forum poops out at around 1MB. Dropbox? Other server company? If you have a piece under 25MB, I can transfer it via email.

Let me know.

Koz

Oh, and as usual, we can’t do anything with a processed piece. Raw work.

Koz

Video is
http://www.acx.com/help/video-lessons-resources/200672590

Yes, noise floor is set -50 when I compress. MAybe should use -59

I get slightly different values between Analyse-Contrast and Wave Stats plugin. Should I trust one more?

Provided that there are no “split lines” or “white space” in the selection, the Wave Stats plugin and “Contrast” should agree pretty closely. If they don’t I’d be very interested to find out why (I wrote the Wave Stats plugin and I’ve fixed several bugs in the Contrast tool).

You got more problems than that. Attached.
The work you sent (first segment) is overprocessed and we can’t take processing out.

Did you save the capture session as either Project or WAV—before processing? I can probably force that segment to work, but the last step before publication, somebody listens to it. That’s where the talking into a bad cellphone or empty wineglass works die.

Wouldn’t you rather listen to that second person reading to you?

Koz

This is where I would sacrifice several small, cute, furry animals to get a crack at the original shoot.

Koz

Arrgh! That clip was cut from my saved “raw” project. I save as a project after editing out my verbal stumbles, and that’s all.
I am emailing you a snip of the out-of-tascam wav file, now. The two files look identical to me, except for the length.
Perhaps there is something like processing happening in the Tascam. I described those settings in first post- 80hz noise reduction.

Steve, the results are very similar, within 10ths. Prob difference between sample and the 29 seconds.

I can HEAR a difference between the two though. I’m going to compare other out-of-tascam vs saved “raw” projects. Something is going on.

Something is going on.

Exactly correct.

The longer work you emailed me appears to be heavily processed. Just MP3 compression processing won’t do that. It’s also muffled like throwing a fluffy towel over the performance. Straight MP3 “hollow wineglass” processing doesn’t do that.

Cast your mind back to the first questions you posted. The sample clips weren’t WAV. They were MP3s and I complained about that. You should know that Audacity doesn’t edit MP3. It converts an MP3 show to is own very high quality internal sound format and edits that. Then it makes a whole new MP3 when you’re done. Each MP3 pass generates more compression sound damage. More gargly, bubbly, honky.

A word on MP3. It’s part of an early video format. Its full family name is MPEG1, Layer 3. It’s a video sound track without the video.

It works by “listening” to your show and deleting parts of it that normal people can’t hear. Many people think all it does is make the show muffled, but it’s much more insidious than that. A top quality violin solo on an expensive violin, under extreme compression, turns into a good solo performed on a violin your 13 year-old kid made in wood shop.

Same tune, same music, very different detail, overtones and harmonics. If you’re listening for it, the performance starts to sound like it was performed into an empty wine glass. If you’re not listening for it, you might miss the damage completely. Which, of course is the object. That and enormously smaller sound files.

I was hoping to avoid doing a hardware analysis, but we’re getting closer and closer. If you just can’t figure out the problem, describe the setup. Details, and part numbers, please.

Koz

Another note. You can’t take MP3 damage out by converting the show to WAV. What that gives you is careful, perfect reproduction of the damage. That’s why we say never do production in MP3. It’s dead-end and you can’t stop it.

A few numbers. The minimum quality for mono is 32, stereo is 64, Audacity default is 128 and anything over that is considered very good to excellent. If you edit a 64 stereo show and export it as another 64, the show damage may be as bad as 32.

Unusable garbage.

This regularly kills people trying to edit somebody else’s existing internet productions into their own show.

Koz

I can HEAR a difference between the two though.

Actually, that brings up another item. You can’t mix, do production or troubleshoot on earbuds, laptop speakers or “computer speakers.”
If you can hold your speakers comfortably in one hand, chances are really good you’re not going to be able to hear half of what we’re talking about when we tell you to listen for something.

If you’re waiting for us to tell you what speakers to buy, you may have a wait. “What Speakers Should I Buy” has been a battle cry since the dawn of recorded sound.

I can tell you that Hollywood fell in love with Sony MDR-7506 headphones. They appear on every sound shoot and everyone looks at you funny if you tell them you don’t have a pair. Of headphones, I mean. I’m not a fan because I can’t wear them for long periods, but they do what they’re supposed to do.

Koz

I’m not.
Did you get the out-of-Tascam file? How is that working?
I don’t know what more to tell you. Tascam DR-05 and pop filter. Rec settings : Format WAV 16 bit (options are 16 and 24-above that are all mp3), 44.1k, Mono, low cut 80hz, rec level 59.
It records on a microSD and then I plug that into comp. I import to Audacity, edit out my mistakes, and then save as a project I note is “raw”. I snipped out a piece of one such and exported it as a wav to send to you yesterday. Then I normally keep working until it’s an overprocessed POS, save it as edited, and then export an mp3 for uploading:)

I got what I recorded today to work with the AUfilter, Normalize, and compress. I was happy no Amplify. Had laptop still today as script on it and no alternative.

Sleuthing results…
When I compare side by side my saved “raw” project and the out-of-Tascam wav, they sound indistinguishable. However, when I export a wav from that project, it changes. That one is reduced, quieter. The levels are slightly different.
This is why the 10 seconds I sent to you, uploaded straight from sd card, sound better than the later clips (“troublesome clip” was edited for mistakes and exported). Most recent clip was opened in Audacity so I could shorten it - still exported. Only the 10 sec samples escaped the import/export. Therein lies the rub.
Here is where I have several choice nsfw things to say.
This means that my most important save is the one as “raw” project.
Is it better, then, to record in wav 24 bit or even stereo wav 24 bit and combine to mono later?
And the question still remains: Can I salvage what I already recorded? Or must I reread? It’s about 8 reading hours - not the end of the world (not insignificant either). Getting it right from here out will be most valuable.

I just got back.

Did you get the out-of-Tascam file? How is that working?

“Yahoo engineers are working very hard to resolve the issue.”

And I believe them because I’m just that kind of guy.

I need to go back through the last messages and review what you’re doing. The worst thing to happen is if the work quality degenerates for no apparent reason.

A pass through Audacity by itself should make very nearly no difference (it’s not perfect because of the two coding shifts). If you can hear a clear difference with no filters or effects, then there is a problem.

As we go.

Koz

Attached a patched clip. The full piece is winging it’s way to your mailbox. The big file makes ACX conformance if you analyze the whole piece.

You have a couple of sharp volume changes in the work which mess with the measurements. This clip was specifically chosen to include one of those emphatic passages. “Try one.” The work won’t pass “peak” without that. Did you lean forward slightly at those words? They’re louder than the others in the original file. It makes compression a little more challenging.

First it should sound like you. Then on to the technical specifications.

I used:

Noise Reduction
Roll-Off
Compression
Normalization.

In general, very gentle settings. I’ll fill in the numbers in a bit. I had one of the tools misbehave and I need to figure out why. Also, It’s possible I’m going to find a formula that doesn’t need one of the tools.

Koz

Can you tell what that is (turn the volume down)? If you can get rid of it, you won’t need the Noise Reduction step.
Koz

Therein lies the rub.

I read that about eight times and I still can’t see the process. Stop me if I say anything wrong.

The Tascam show plays OK from Audacity and the Saved Audacity Project plays perfectly when re-opened. I assume the show also plays OK at the Tascam headphone connection? It should also play OK in QuickTime Player?

When Audacity Exports a WAV of the work (no filters or effects yet), the file sounds horribly damaged and muffled when opened in a fresh Audacity? This is the step I can’t tell where the damage is coming from. On the other hand, if the exported WAV is exactly the same show just a little off volume, then yours works pretty much like mine does, although I’ve never tried opening the same show in both formats.


If you tested the long file I sent, did you get compliant peak, RMS and noise values? The RMS isn’t going to meet by much. I want to make sure you get the same numbers I do. And yes, given that quality of original work, I think it’s not that hard to meet guidelines.

It sure would be good to not have to fight that air conditioner or ventilation fan or whatever that is.

Koz

Long Cookie Segment Patch from Original RAW Capture.

Noise Reduction 1
Drag-select 21.2 seconds to 24 seconds. Effect > Noise Removal > Profile. It’s super important that you don’t catch any of your voice or breathing in that sample—or as little as possible. But you still need to give Noise Reduction enough to “chew” on.

Noise Reduction 2
Select the whole piece (Click above MUTE).
Effect > Noise Reduction:
9dB
0.00dB
150Hz
0.15Sec

=================
Apply Steve’s custom LF_Rolloff filter. Attached. Double click to get the XML file from the ZIP. Effect > Equalization. Import the custom XML file.

http://manual.audacityteam.org/o/man/equalization.html
Scroll down the manual to importing and exporting. The special filter has to be applied with the “Length of Filter” slider all the way up. Attached Illustration.
This is another one of those struggle through installing it once and you just apply it from then on.

=================
Compression:
-20dB
-40dB
3:1
0.2 sec
1 sec.
OK.

===============
Normalize
Effect > Normalize: [X]Normalize to -3.2 [X]Remove DC

That’s it. The work should pass all three measurements. If it falls below RMS 23dB, UNDO your way back to compression and try 3.5:1 or 4:1.

Koz
Screen Shot 2015-02-20 at 21.16.50.png
LF_rolloff_for_speech.xml.zip (326 Bytes)

All of those settings are very gentle and a primary reason the voice at the end sounds startlingly like the one at the beginning. Note you can still hear very slight noise in the background if you’re looking for it. But it does meet noise specification and it didn’t damage your voice.

And then I took another cookie.

Koz

I was like WHY are you on about cookies??? Totally baffled… oooohh. hahaha