RE: nintendoeats, high frequency game audio processing problem

RE: nintendoeats, high frequency game audio processing problem


Hello nintendoeats… happy to help out a fellow game audio pro.

So if anyone not from happens to be viewing this, what happened is this guy was trying to filter out dialogue EQ with brutally excessive sibilance that could not be filtered out. Gating was also not an option as it overlapped the dialogue which appears to have caused it to begin with. So even when he placed an absurdly aggressive low pass at 1k with a 100db poll, he was still getting pounded with abrasive 7-20k sibilance. It just would not go away.

The first thing you need to do is see exactly where these frequencies are occurring and at which respective time points. In the second picture below, the white streak at the top is where your 1K filter failed to attenuate static like white noise. My guess is this is because your filter plugin was not equipped to address frequencies over 20k. My monitors go to 30,000 hz, and the spectrum shows that HF information increases instead of decreases as it approaches 20k. That’s how you can infer this. You can also see from the spectrograph that the audible problem sibilance is occurring at 5-10k. Though your filter practically a cliff, it still didn’t work and left that range remaining. My first step was to engage calibrate the repair parameters and engage the instant process feature on RX advanced. I took care of the 16-30K bleed first.


Now its gone.

Knowing that the human ear perceives the vowels in the mid 200 up to about the 800 range, and consonants in the 3k range, I targeted the sibilance here with the lasso tool. What this does it it attenuates is it allows you to bring down frequencies only in one range, only at one time. This is impossible to do in a normal DAW with a music mixing plugin. Below you can see those specific frequencies have been removed.

Same to the first cluster.

Gone. V V

As I move forward you can see the progress log build. To undo you simply click on the previous macro.

Now (below) for the middle cluster, I had to change the before and after weighting. This is not an EQ plugin. You not ‘attenuating’ frequencies. You are erasing them and having them artificially re-constructed based on the information before and after the lasso’d cluster/blob. The “Band” parameter effects the resolution. This is how many sub-divisions of the selected region are made when the replacement audio is re-synthesized, and the “Direction” determines how far above/below (vertical) or infront/behind (horizontal) the plugin examines to determine how to synthesize the frequencies algorithmically.


And here is the last section that had noticeable sibilance.


I feel rather awkward saying it, and I do appreciate the help, but I think you have written a solution to the wrong problem. The file you posted still contains the high frequency noise. All I can hear different is that you have very strongly de-essed it. Also, while that lasso tool looks very cool, it isn’t terribly useful to me since I don’t have access to RX advanced.

Just to confirm that this is not an issue with Audacity’s EQ and Low-Pass functions, I downsampled the file to 8KHz and the noise remained.

So that there is a copy in this thread, here is the problem WAV


@nintendoeats, hey man. Glad you followed up before that last link was deleted.

Let me give this another listen. One sec.


Ok. Here’s with 12-20 removed. I guess you weren’t talking about the sibilance on the very top?


Nope, It’s still there. I’m talking about the consistent tone which runs throughout the entire clip. I wish that I could circle it and go “no that!” but if I could that then I would probably also be able to remove it.

I’ve listened to this clip through 2 high quality pairs of headphones and some pretty good speakers. It’s softened quite a bit by the speakers, but through both sets of headphones this sound is all-consuming to me (probably because I’ve listened to the clip 100 times trying to wipe it out).


Ok. Are you looking at the cross between where the “T” of cut crosses with the “Ow” of out?

There’s really no “I” vowel on the “It’s”…right? So why not just go after the sibilance on the “T-S” of it’s? In that one isolated area? You don’t need RX to do that…

In between 0 and .76 can you pinpoint the most prominent example of what you’re trying to axe out?


Looking at in what sense?


This one is a wider sweep, now that I think I have a sense of what you’re after. See if this is closer…if I can narrow it down I can give you some ideas on how to fix it.

It may not be gone but see if its closer.


I would say that it is most obvious over the “tow”, 0.31 to 0.63. It’s REALLY weird to me that you don’t hear it though. I think we may have slightly different effective hearing ranges, because It’s really loud to me.

I do have a thought; if this one sound has components everywhere from 1KHz to 15KHz, that would explain why no high-pass or low-pass entirely gets rid of it. It sounds very high pitched, but that’s also the range where it is most likely to be heard. What remains after a 1KHz low-pass filter may only become clearly audible when most of the rest of the sound has been removed. Further, since I HAVE heard this sound countless times, it’s possible that my brain expects the higher frequency components to coexist with the low frequency ones and tries to fill them in. Just a theory, but it’s the only explanation I have for why such a noise would continue to be heard after a huge low-pass filter is applied.


Nope, still entirely present in that last clip.


I have to go for the moment. I appreciate your help, but I think that trying to remove a sound you don’t hear might a good way to drive yourself nuts alongside me. I don’t NEED this clip, I only asked the question because it bothered me that I didn’t understand what the problem was. I would feel bad to know that you spent this much of your weekend on something that is ultimately frivolous.


Ok. Gotchya. On that last clip I just snagged the sibilance at the beginning and end. I gotta work on something else, but I’ll follow up later tonight…now that I know what you’re hearing, let me try something else when I have a sec later.


Oh, I certainly heard that, but I didn’t realize .31 to .63 was the one you were one you were mainly concerned with as I found them to be far more distracting lol. If you look at those gigantic orange blips at the beginning, those are the ones I thought you were trying to get rid of.

Once you pointed it out, I was able to isolate it.

This is not a usable file but I did quickly isolate it once I realized which one you were talking about. I believe what I did locks down the location of the problem frequency and brings them under control, but it does some pretty severe damage to everything else in the process.

The thing is living in this region right here (which is spread between 5.5K - 8K), and it is one stubborn ass motherfucker to get rid of.

You still hear a little dabble of it (and you can see it in the faint blueish black on the spectrograph). If this was my job, I’d find a compromise because I’m certain at this point anything do to cancel it in its entirety is going to make the phrase unrecognizable. I would deal with the sibilance before and after, cut my losses and just pound that one isolated portion from .31 to .38 with a multi-band attenuator.

I think a multi band attenuator is the best tool for this if you don’t have a spectral editor. If you’re on a time-based DAW like audacity or Protools I would automate the attenuator to kick the shit of this thing, but only in that one specific spot, then deactivate it at .38 sec so it doesn’t damage everything else.


To your original question of why are you still hearing despite that 100db filter and why won’t it go away?

I’m pretty sure its because filters, de-essers/MB-attenuators, and algorithmic cancellation processors are not designed to complete eradicate rouge frequencies in the same way that selecting and muting a portion of a file will make something completely go away.

This was certainly an interesting challenge to run into! Glad you asked :slight_smile: