Help with LF_Rolloff_for_speech

Hello, forum folk! I need to revisit an old question regarding the quality of my audio sample. I got great advice and have used Noise Reduction and Normalizer with success on the first four audio chapters of my sci fi novel. I also, from that same advice, downloaded the LF file referenced above, but I did something wrong, because it has not subsequently showed up on the Noise Reduction page where the forum advice said it would, as an option (as I remember–I can’t find the thread now!). I’m pretty sure I unzipped it right.

It may be where I put it after I downloaded it. I put it in my Audacity data files, in documents on my hard drive. Maybe I was supposed to put it in the program file?

If anyone knows what I mean, perhaps you could help me to make this feature operate? Thanks!

LF_rolloff is a plug-in for Effect > Equalizer. When you get it installed, you should use it with the Equalizer Length slider all the way up.


Adding Equalizer Curves
– Select something on the timeline.
– Effect > Equalization > Save/Manage Curves > Import
– Point at LF_rolloff_for_speech.xml > OK. (it won’t open the ZIP. You have to decompress it)
– LF Rolloff for speech now appears in the equalization curve list.

Use LF_rolloff early in the filter process. If you have low-pitch tones or rumble (more common than you think), it can mess up the other tools. One of the filters or tools like compressor may be trying to process the barely audible rumble instead of your voice.


But how do you install it? Where do you save it to?

You can put LF_rolloff_for_speech.xml on your desktop. When the Equalization installer asks you where you put it, point to the desktop. Instruction #3 in bold.


Oh, Hi, Koz! The instructions in bold, can you put the link? This is probably where I messed up! And when you say ‘point to’ do you mean, ‘click on’?

And when you say ‘point to’ do you mean, ‘click on’?


The instructions in bold, can you put the link?

It’s not a link. It’s the instructions I posted in bold text five messages ago.

Adding Equalizer Curves
– Select something on the timeline.
– Effect > Equalization > Save/Manage Curves > Import
– Point at LF_rolloff_for_speech.xml > OK. (it won’t open the ZIP. You have to decompress it)
– LF Rolloff for speech now appears in the equalization curve list.

Okay, by Equalization Curve List I think you might mean Select Curve? If so, LF_Rollout does not appear in the list, but something called unnamed does (no capital). Is that it? Should I rename it as suggested by a pop-up dialogue list, LF_Rollout? Thank you!

Wow I think it installed, after I quit clicking the LF_Rolloff file on my desktop and clicked on it instead in the dialogue box that came up when I clicked on Import.

The file is now on the list! I selected some audio with the Length of Filter all the way right (up), and LF rolloff for speech selected. I clicked on OK. Something happened. I tried to take a screen shot of the Equalization boxes, but couldn’t. May I describe the before and after curves instead? Before I clicked OK and began the Equalization, the curve began in a straight line at the intersection of -18 dB and 20 Hz, ran along a straight line to 59Hz, and then rose steeply, like 85 degree angle to 0dB and I forgot to look if it ran on in a straight line after that. After I (think I) did the equalization process, the curve began at -30dB and 20Hz, ran along in a straight line until 59Hz, and then rose to 0dB and 100Hz, which looks almost straight up, again at perhaps an 85 degree angle, and after it got to 0dB, it ran along in a straight line to the edge of the dialogue box.

Is this the intended result of doing something with the LF rolloff for speech file? I could not hear any difference in the audio before the selection and the selection, but I am getting a little deaf.

Thank you!

For the most part you don’t hear this filter running. It’s intended to suppress very low pitch sounds like thunder, rumble, or earthquakes. Sometimes microphones make these sounds by accident, but none of them are welcome.

You can get similar results with a normal Audacity tool.

Select the whole unprocessed clip or show by clicking just above MUTE.
Effect > High Pass Filter (set up according to the attached) > OK.

I also included two pictures of Analyze > Plot Spectrum. Click on the picture if you need to in order to see the whole thing. The first one is a natural voice performance without filtering. Everything to the left of 20Hz can not be heard but can create processing problems. The next analysis is after LF_rolloff. Much of the stuff to the left of 100Hz is missing. 100Hz filtering is use in many voice performance recordings. I have it built in to several of my sound mixers.

Screen Shot 2015-06-10 at 18.49.05.png
Screen Shot 2015-06-10 at 18.48.37.png
Screen Shot 2015-06-10 at 18.43.17.png

Koz, could you please look at my graph in the file attached below–does it look okay? That is, within parameters for an audio book?

I learned how to take a screenshot, but it would not paste here, I had to save it to dropbox first.
Screenshot 2015-06-12 10.03.27(2).jpg

Koz, could you please look at my graph in the file attached below–does it look okay? That is, within parameters for an audio book?

Thank you for the graphic. We can’t tell directly from the graphic. That’s just a confirmation that your corrections seemed to go OK.

We have to run tests on your sample clip.

I’m fascinated that your spectrum analysis doesn’t look anything like mine. It’s rough to tell you to make corrections when we’re not looking at the same thing. I need to go away for a while. I’m typing this on a dead run on the way out to the car. Are you using Audacity 2.1.0?


Yes, I just checked again to make sure (I checked once before when Gale asked me the same question), yes, it’s 2.1.0 and I just downloaded it May 16. I wonder if it’s because I use Vista (need it to print one of my pattern generating software products)?

Well, I am sorry you have to ‘go away for awhile’ (that’s what we used to say about Uncle Clyde whenever he robbed a bank) and hope you will be back safe and sound soon! Be careful! The graph can wait but I hope you won’t forget!

I’m back from the bank. I mean I’m back from shopping.

I figured out the graph. Note the Size setting on yours is 512 and mine is 16384. My setting lets you see more of the purple pattern below (to the left) of 100Hz. That’s where Steve’s LF_rolloff filter does it job, so it’s good to see it working.

As I said somewhere up the thread, humans can’t hear below 20Hz and most rumble and low pitched voice interference happens from 100 Hz and lower. Steve’s filter greatly reduces everything below 100Hz. It doesn’t just chop it off because you may be able to hear that working, so it takes interference out gently but firmly.


I think you were trying to send a “docx” file to the forum. That’s a Microsoft Windows proprietary document format and I’m pretty sure the forum won’t accept it. What you’re supposed to do is snap a picture of the screen (PrtScn button? It’s been a while) which copies it onto your clipboard. Then you open MS Paint and paste the picture into that. Make any changes you want (cropping, colors, text) and then Save it as a JPEG picture file. The forum will accept those.

Now when you produce a Plot Spectrum before and after LF_rolloff, you should see Steve taking out the purple pattern to the left of 100Hz.

I don’t remember if I sent you flynwill’s analysis plugin. Select the whole clip by clicking just above MUTE. Analyze > ACX Check.

That will tell you in one sweep whether the clip will pass ACX automated conformance testing.


Yes, you did say it before, but I didn’t and don’t understand it. Is 100Hz ‘below’ or above 20Hz’s? It’s calling sounds ‘lower’ and ‘higher’ that mess me up. In numbers, 100 is higher than 20 but maybe not in this context. Maybe it’s lower. And it wouldn’t communicate anything to me anyway. What I need to know, initially, is what it does to the listener’s perception of the audio. If we can’t hear below 20Hz, why cut off those rumbles at 100Hz’s? Presuming they’re lower? See my confusion?

Also, Print Screen as it turns out depends on several factors, the keyboard brand, the operating system. In any case, since I successfully uploaded, using Dropbox worked for my combination.

I can’t find a mute button anywhere. That is a very useful tool, and you also mentioned that before but I couldn’t find the mute button then either, and with so many other challenges (which have been met, thanks to you and the kindness of others on the forum) I had to let it go. I will google ‘mute button Audacity’ and see if there’s a guide anywhere. It must sound stupid, but I’ve looked at each button and tried to click on everything that looks like a speaker, mute usually being a line through it, but nothing emerges from the dense array.

Communication is so difficult. Like, I should have said we used to remark, in my family, that Uncle Clyde was ‘going away for awhile’ after he’d been CONVICTED of visiting the bank. How about a new beatitude, Blessed are the careful communicators, for they shall sit right in the lap of God drinking vino.

Oh, I googled mute button and now see it quite clearly, the word Mute on the line with the track, not up in the toolbars. Ok now the truth will out: audio book quality, or not? Thanks!

Dear Koz, I was able to complete the Analysis/Contrast step, but I do not have anything called ACX check and it’s probably because I never got that file to download. It does pass the WCAG2 test. 49.9 dB Average RMS–that’s the difference between the background noise and the speech.

I read through a thread with you and Santina724 and hardly understood a word, but did notice that the numbers you gave for that person’s Noise Reduction Step 2 were different than you gave me for mine, which were 12/6/6 with the level all the way to the right on that last item. Did you give me those numbers based on the sample file I uploaded, and will they be good for all my chapters as long as I don’t change the room, mic, etc.? Also, may I ask, does the number of times I do any noise reduction reduce the volume? I have been doing this, if I found some breathing noise or whatever, I just selected the whole file and ran the noise reduction for that sound (rationale, it would find all the times I took a breath like that one and I wouldn’t have to look for them and manually delete them) but I think I reduced the volume. The peaks on the file seemed to get smaller anyway. Should you just run noise reduction as seldom as possible, only for typical, not atypical, noises? And find the others noises and manually delete them?

Anyway, I didn’t test all my finished chapters, the four of them, but the one I was checking I think failed the loudness test. It was 3.1. I don’t know if ‘passing’ the WCAG2 test includes loudness, so if not, could you tell me what number to put in the New Peak Amplitude box in Amplify, if there is one?

Right now this is the routine: I’m recording, then applying Steve’s magic LF file, then Equalization, Normalization, and Noise Reduction (several times on that last step), not necessarily in that order, plus I tighten up any pauses–takes a while, that last step. I will add that ACX check when I have the file (I’m gonna go look for it), and meanwhile will do Analyze/Contrast and see if each file passes the WCAG2 test at least. So I’m asking, is this routine okay? You and Santina talked about so many other things!!!

Thank you!

I need to read through that carefully to catch all the points. Yes, there’s nothing at all simple about this and it does use its own language.

Remember that oboe note at the beginning of the orchestra? Thats 440Hz (normally). High-pitch baby screaming on a jet? That’s around 3000Hz. Fingernails on blackboard is even higher, say 5000Hz. Thunder, earthquakes and bass violins live around 50-200Hz. Actually, earthquakes can be even lower than that, so low you don’t even hear them. You know the sound is there because it shakes the bookcases.

There are some microphones that produce thunder tones by accident when you record. The computer sees these tones as a part of the show and they take part in sound processing whether you want them to or not. LF_rolloff suppresses them.

I need to go do weekend things for a while.

You can just about see the question marks over my head.

Web Content Accessibility Guidelines (WCAG) 2.0 have to do with building a web page or display system so the most possible people can use it. Did someone say that was a sound test?

OK, anyway, I enclosed flynwill’s sound analyzer, acx-check.ny. He programmed it in Nyquist language, so that’s the “.ny” at the end of the name. He didn’t write it in New York. Windows may or may not let you see the .ny when you download the file.

This is where it gets fuzzy because I’m not a Windows user. The Audacity application has a folder called “plug-ins.” Drag acx-check.ny into that folder (example picture attached) and start or restart Audacity.

Open Audacity normally with one of your chapters or performances. Select the whole thing by clicking just above MUTE.

Analyze > ACX Check (example picture attached. ACX Check is fourth down.).

As I think I said before, this is simple steps once you have everything installed. Getting it installed is the pages of instructions.

Once you get everything settled, post a sound segment what you and ACX-Check thinks passed and I’ll run it through my tests here and we’ll see if we agree. If it’s much over about fifteen seconds, you may need to post it through DropBox.

Screen Shot 2015-06-13 at 18.50.26.png
Screen Shot 2015-06-13 at 18.39.41.png
acx-check.ny (5.6 KB)

I’m stuck on Step One. I can’t seem to download the acx-check file. The computer gives me two options, it asks me what program I want to use to download it, or do I want to use a program on the computer already–Browse. Neither option works for me. Microsoft (the web option took me there) suggest Nyquist, in the Audacity folder, so I tried that one, but the error message said No Items Found in Folder, and there was nowhere to go from there. I tried Browse and saving it directly to the Plug-In folder, but again got the message No Items Found in Folder. I didn’t try just browsing for the Download folder, because the computer seems to want a program to open the file, not a folder to download it in.

Is there any tutorial to download it?

Regarding the WCAG ‘test,’ that’s a result on the Analyze/Contrast screen, which collects sound data and then gives some numbers about background noise compared to foreground, so one might think it’s a sound test. Newbies might.