Mountain River Hiss. Best way to remove?

Hello! I’m recording in a natural environment where there is a river about 500 meters away. This is a sample of the hiss sound I’m picking up. Normally I will be increasing the volume of the recording, so it ends up being much louder than this, but I believe we are supposed to post unchanged samples.

I have been using noise reduction effect, but I was wondering if there was some other magic out there that would be more appropriate.

If it matters, I’m using Audacity 2.3.2, a Focusrite Scarlet 2i2, and a FiveO Montarbo PM-85 dynamic mic. I’m not in a position to change equipment right now. I’m interested in capturing the other natural sounds, so isolating the river noise pre-recording it’s my favorite (or best) option.

The actual thing I am recording is human voice. I increase the volume using the RMS normalize plugin. I’m also sometimes doing a base cut with the Equalizer.

Thanks!! I’m really a noob when it comes to processing, so please talk to me like I know nothing.


It’s spectrum & low volume (<-60dB) lead me to suspect that hiss is the inherent noise of the electronics,rather than a river.
Try an experiment: if you record with the mic unplugged, any hiss captured cannot be a river.

That’s interesting. I guess because I can hear it with my ear directly I assumed it was being picked up by the mic.

I’m guessing that what you asked for was for me to unplug the mic from the Focusright mixer while leaving that mixer selected as the input in audacity. That is what is above.

Do you think it is just system noise?

I read that three times. What is the goal? Are you trying to announce with the river in the background? If you need “clean” river sound, you should be at the river recording it and not a third of a mile away.

While you’re announcing in your quiet, echo-free room, can you get the Scarlett knob ring to change colors?

The gain controls have concentric tri-colour LED ‘rings’ to confirm signal level: green indicates an input level of at least -24 dBFS (i.e., ‘signal present’), the ring then turns amber at -6 dBFS to indicate that the signal is close to clipping, and finally to red at 0 dBFS (digital clipping).

You should be able to get an occasional amber. That corresponds to the -10dB to -6dB recommended for live announcing.

This is where you ask us where to set the Scarlett volume control. You set it wherever you can get those colored lights to come on. Dynamic (moving coil) microphones tend to have “gentle” volume and nobody would be shocked if you had to run the volume control all the way up…and still not have enough.

It’s not unusual for dynamic microphone users to have a Cloud Lifter between the microphone and the interface to provide more volume. I can’t find the example for that right this second…

Yes, the electronic noise in the system is the addition of all the analog noises together, but the most important noise maker by far is the microphone preamplifier. It’s the first thing the microphone sound hits when the sound enters the Scarlett and the overall system noise will never every get better than that. That’s what they mean when sound mixer makers brag about having terrific microphone preamplifiers. This is from a Mackie sound mixer spec sheet.

Screen Shot 2019-11-18 at 19.02.35.png
It almost doesn’t matter what the rest of the mixer is doing. That one thing is going to determine the quality of the recording.

There is one oddball thing you should pay attention to. Microphone preamplifiers change noise depending on whether or not you have a microphone plugged in. The two are matched to each other. If you just unplug the microphone, you could start picking up the local radio stations. I’m not joshing.

So instead of unplugging the microphone, wrap it in heavy towels or other sondproofing. That will give you a much better idea of the actual noise. You’ll get to a point where no matter what you do, the noise stops decreasing. That’s the real electronic noise.


There it is. The Shure SM7B. Terrific microphone, but “restrained.” It’s frequently sold with a companion Cloudlifter.

Screen Shot 2019-11-18 at 19.23.45.png

That’s what The Joe Rogan Experience uses.

Screen Shot 2019-11-18 at 19.29.18.png
I don’t know that he’s using a Lifter, but I wouldn’t be surprised.


I’m so sorry. I thought the subject line made it clear. I want to remove the river hiss. I am living in the forest making recordings in a small cabin. The podcast voice content is such that having bird sound and other forest sounds is a good match, so I don’t mind having those. And in my situation I couldn’t eliminate them anyway, so I feel it is better to just let them be included.

However the water sound, to me, adds no value because it is just a constant hiss sound. I don’t have any perception of it being a river, just an unchanging shhhhhhhhhhh sound. So I wanted to try and remove it. When I hear it directly through my ear, (not through equipment) it seems quite loud, but that could just be because things are naturally quiet here.

I thought there might be something more effective/less destructive than the Noise Reduction effect. But I can continue to use that, of course.

I was actually going to do a different post about getting a greater output from the Scarlet 2i2 because I thought that was a different issue. In brief, I do get amber at the peaks. However clipping starts at .6 which I don’t understand, hence why I was going to do another post.

I will try the wrapping of the mic in cloth and upload another clip. I realize that just listening to this clip there is almost nothing. But once I run the RMS effect, the background hiss on the tracks with my voice is much louder. Should I upload an example of that?

Thank you so much. I feel like you always give 1000% here and I appreciate it.

Limiting to ~0.5 (-6dB) can be due to recording in mono from a stereo device,
see … Recording in mono severely limits the waveform - why? - #3 by steve
If you set everything to record in stereo you should lose the ~0.5 limit.

Limiting to ~0.5 (-6dB) can be due to recording in mono from a stereo device

What he said.

Since you seem to be doing everything right, record a simple voice sound test and post it here. We may be able to help with a specific case.

There was a recent change to this process. If you don’t get the 2019-11-15 version, reload the page.

Quick note. You’re not supposed to go “Diving For Noise.” It doesn’t count when you play a clip and turn the speaker volume all the way up. Most people’s voice tracks are going to fail that test.

More later.


Wow. The info in this post changes everything for me.

I think I could even say the problem is resolved. The hiss seems negligible and when I do the noise reduction I’m very happy with the results.

If anyone could listen to this test and offer any other suggestions, such as a better way to remove any remaining noise that would be great.

I did have a question though about the info in this post:

Following those instructions, I record in stereo, then I split the tracks, move the track with audio in it to pan center and delete the other track. Is there some setting I should change to avoid those steps? I’d even be content if there was a way to get a macro to do it.

Windows 10 Home
Audacity 2.3.2

Thank you so much! This is a huge help.

Is there some setting I should change to avoid those steps?

Menu on left > Split Stereo to Mono. Delete one track. Mono automatically plays to left and right headphones or speakers.

I’d even be content if there was a way to get a macro to do it.

The Macro’s worse. Tools > Apply Macro > Select the Macro. Or Tools > Apply Macro > Select the Macro > OK.

Is 2019-11-20 before or after you applied Noise Reduction?


OK. I got it. I think your clip is broken. If you carefully listen to the background noise before and after you speak, they’re different. Further, after “Amazing,” the background noise changes for two seconds. I can’t tell if that’s you breathing after the words…?

You can’t do that. The processing tools will never be able to keep up with that. You should have the same background noise all the time. Read for the full eight seconds. Read the milk carton/cereal box. You can announce as long as you want, just cut it so the total never goes over ten seconds. We’re not trying for a theater award. We just need a lot of different words so we can tell what you’re doing.

You’re too close to the microphone. You sound like you’re trying to talk right into my ear. I can feel your breath.

Either a Hawaiian Shaka away from the microphone, or a power fist away from the pop and blast filter (if you have one).


And yes, I can do some serious noise reduction with the normal tools but not with that first clip.


There’s a test clip trick.

Split Stereo To Mono first thing. If you send the forum a mono clip (one blue wave), you can talk for 18 seconds.


I’m not familiar with the term “clip is broken”. I’m not being snarky, just not sure if this is a technical term, or you just mean something is wrong. When you say clip, do you mean the track I uploaded, or are you talking about clipping? EDIT: OK, you must just be saying something is wrong. Any idea what I did to break it?

I don’t know why the background noise would be different in different parts. Personally I can’t hear the difference, but I am new at this.

Sorry for getting in your ear. :confused: I have been trying to keep the mic close to reduce room echo, but yes, the mouth noises are not good.

As per instructions, I have done nothing to any of the clips I uploaded.

Here is a new recording. Gain is all the way up on the Focusrite. I’m speaking directly into the mic, one haka away.

Same setup but I am outside. I realize I am shouting a bit.

I was also not clear if the two seconds of breath holding was to be included in the clip to hear the background noise, or if it was just to prevent the sucking of air before speaking.

A friendly suggestion… Why don’t you put the exact sentence you want people to record? It might save everyone some time.


“clip is broken”

That’s a combination phrase. A “clip” is a segment of a performance taken from the action of attacking a piece of audiotape with a razor blade or scissors. “Broken” refers to a sound clip that is not suitable for purpose, to borrow a British phrase.

Personally I can’t hear the difference

It may not be obvious, but it sticks out as damaged background sound when I apply the reduction tools.

If you’ve been using Noise Reduction, you know that the process has two steps. The Profile step where you select some representative noise and let the tool “sniff” it to know what to reduce. Then the actual application where it goes down the performance looking for a match to the profile. In the case of that last submission, the noise before and after the speech didn’t match and only one got successful noise reduction. If that happens a lot, it might be one problem with some of your adventures in noise reduction.

Any idea what I did to break it?

It’s possible I’m listening to you breathing a sign of relief after a successful presentation. The character of the sound is similar to the river and the electronic noises, just louder.

I have been trying to keep the mic close to reduce room echo

There’s nothing magic about the plastic pipes. You can use branches and heavy blankets. Surround the microphone on five sides with acoustic dead walls, without having to soundproof the whole room. One warning: Light, airy, fluffy comfortable blankets need not apply. I have a large square sleeping bag which would probably work, but it’s hard to beat those cheap moving blankets.

As per instructions, I have done nothing to any of the clips I uploaded.

Yes, that was obvious after I started to work with the file. We can’t take effects, filters and corrections out of a sound file. If you submit a corrected or processed voice, we’re at full stop. We may be able to tell it’s broken, but not why. That’s not useful.

Why don’t you put the exact sentence you want people to record? It might save everyone some time.

[Writing that down]

Actually, I did once write a piece about “Milk produced by contented cows in the barn at Catskill Farms Dairy in New York.” Completely fake. I should see if I can find that again.

As we go.


Screen Shot 2019-11-19 at 20.28.12.png

I don’t suppose you can do all your recording outside?

With a little stiff processing, it meets audiobook technical standards. First three values and the sentence 2/3 down.

Screen Shot 2019-11-19 at 20.39.42.png
That, and the voice is much better without that recording-in-the-kitchen sound.

I don’t remember from earlier. Are you going for audiobooks, or personal podcasts? Other? That outside clip with the birds and water in the background is pretty much perfect.


Oopse. Wrong Polynesian culture, wrong word. A Maori haka really wouldn’t be helpful for the recording. But here’s a haka with a shaka:
19-11-20 10_24_43-haka - Google Search - Opera.png
In any case, I was keeping the mic a hand span away from my mouth, aka a shaka.

WOW. That’s amazing. Please teach me your magic. :mrgreen:

Personal podcast. It’s not a commercial audio book, so the bird sound is not a problem. It’s actually a good fit with the show.

And yes, I can do all of these recordings outside. Once one starts recording, though, the changes to the forest become much more obvious. After lunch the birds go crazy. By about 4 the crickets start to chime in and around 6 the frogs join into a wonderful symphony that is too much for background noise. And the rain anytime is too much.

But yes, outside is where I will do this project.

I’m a fool for not asking for your help six months ago. If you can tell me the processing you did that would be amazing. The only effects I have worked with in the past are noise removal, RMS Normalize, Normalize, amplify, and some equalizer. But really I don’t know very much at all about these things.

I found audiobook mastering handy because their standards are spittin-distance of broadcast and if you can make one you can probably make the other. Once you get it that far, you can submit to anybody else with success.

In general, I applied published audiobook mastering followed by stiff noise reduction and then DeEssing because your microphone has a tendency to punish “SSS” sounds. That last step maybe why your mastered work sounds more “civilized” than it did earlier.


This is the writeup on mastering. Besides the actual tool usage, you should know you need to install RMS Normalize, and, if you use it, ACX Check. That’s where I got that INFO panel from.

Scroll down to PROCESS. Please don’t ad-lib. Each of the tools depends on the tools before it. It’s a “suite”, a harmonious grouping.


I drag-selected just less than a second of nature sounds before you start talking and Effect > Noise Reduction > Profile. Be careful not to include any desired sound in that sample. This is what I complained about earlier. That earlier sample had (by my count) three different background noises. You can’t do that.

There’s a funnier version of getting the Profile wrong. Include your voice by accident and Noise Reduction will try to remove you.

Then Effect > Noise Reduction: 9, 6, 6 > OK.

You can watch the rumply trash sound in the first second of the blue waves decrease. The sound doesn’t really vanish. That’s an artifact of how the blue waves work. But it does give you a sense of completion.

But wait. There’s more.

If you then listen to the voice you will find harsh, piercing SS sounds. That’s Essing. There is a de-esser tool that has to be installed.

These are the settings I used.

Screen Shot 2019-11-20 at 3.59.16.png
Steve (forum administrator) has a DeEsser as well. Since this is the last tool in your correction suite, you can experiment a bit. Steve’s was designed after I got Paul L’s to work, so I have no experience with it.

At least in the case of Paul L’s DeEsser, it’s an absolute tool. To get the results I did, it must be used after the other tools. It will not automatically trip down the show looking for Essing on its own. It has to be spoon fed with mastered audio.

And that’s how I did it.

You have a terrific announcing voice after we got you out of your echo chamber, and given the show concept the birds in the background is a natural—at least until it starts snowing.