iZotope RX 7 Tutorial ② Removing Breath and Various Noise

Author: sleepfreaks

Demonstrating Commonly Used Modules

RX 7

In our 2nd tutorial on iZotope’s audio repair tool RX7, we will be looking at editing using various modules.
Though many modules are included, we decided to focus on what we feel are the most commonly used ones.

iZotope RX 7 Tutorial ② – Video Access

Breath Control

You can find the RX7 modules in the menu on the right.


You can choose what to use from here, so lets try out Breath Control first.


We have a sample here with obvious breath noise so lets hear how it sounds.

We will be applying Breath Control to this sample.
Its controls are simple; use Sensitivity to setup the sensitivity for breath detection, and use Target level to determine how loud the breath will be.
We’ll start by setting the target level to -inf and adjust the sensitivity. Press the Preview button and play to hear the effect in use.


For this sample, even when the Sensitivity is set to the the breath part and normal talking voice manage to be analyzed separately.

If volume reduction is heard on non-breath parts as well, adjust the sensitivity to where it just barely mutes the breath.
If you don’t want to completely remove the breath, adjust the Target level to determine the volume. Though the target level will give you the level shown, you can switch to Gain to get a reduction in level/volume relative to the volume of the breath sound as well.
If the preview sounds good, click the Render button.



Next please give this sample listen.

There is lots of various noise here. Lets try cleaning up this audio sample.

First off, that loud zap sounding noise was troublesome. We’ll use De-click to get rid of this kind of sound.


  • Algorithm:The type of analysis used to remove clicks. We recommend previewing the sound while choosing which type to use.
  • Frequency skew:Determines frequency band for click detection. Lower frequencies are detected on the left, high frequencies on the right.
  • Sensitivity:Determines click detection sensitivity. Raising it too high leads to unnatural results. It’s best to find a balanced point.
  • Click Widening:Determines the time of removal applied to a detected click sound. To fix rapid click sounds and other issues from digital waveforms, etc. we would set this setting a little higher, but this will also cause issues when set too high. Find the best point for this setting as well.

For this sample, we used MULTIBAND (RANDOM CLICKS) with a focus on lower frequencies to get a clean result.

Spectral De-noise

Next lets try getting rid of the white noise in this sample.
We will be using Spectral De-noise to do this.

spectral de-noise

Spectral De-noise analyses unpleasant frequencies and subtracts them from the frequency to remove noise.
It’s best to select a portion containing just noise to analyse the sound.


  • Threshold:Lowering it gives a more natural reduction but may often result in plenty of left over noise. Raising it gets rid of more noise but it can also get rid of the sounds you want to keep in the process.
  • Reduction:Refers to the amount of noise reduced, and works hand in hand with threshold, making a balance between the two parameters crucial.
  • Quality:You can keep the Quality at Best if your CPU can handle it.
  • Artifact Control:The lower its set, the better the separation between the noise and the voice. However, it can result in a loss in vocal quality. Raise it to the right up until a point where the vocal quality loss is not noticeable.
  • Reduction curve:By ticking the box, a blue line will be displayed in the center of the analyser. You can adjust the strength of noise reduction by frequency using this curve. Click to create a new point, and raising it will lower the noise reduction on that frequency, while lowering it will raise the noise reduction.
  • Smoothing:Move this slider to the right to smooth out the Reduction curve.

The noise has been greatly reduced.


Next, we will look at the first “sa” sound, where we have an explosive air sound and try to remove it.
To do so, we will look at the De-plosive module.


  • Frequency Limit:Determines the highest point for plosive sound removal. Move to the left if it begins to cut into the low frequencies on your audio.
  • Sensitivity:Determines the sensitivity of plosive sound analysis. The higher its set the better it can detect plosive sound, but can interfere with the vocal sound.
  • Strength:Determines how strong the removal effect of the plosive sounds are. If set too low you won’t remove any plosive sound, and the higher its set the more reduction on the plosive sounds. However, if set too high the overall performance will lose it’s low end.

By adjusting accordingly, we are able to remove the plosive sounds without cutting into the low-end of the voice.


As a final touch lets remove the Ess sounds. We will be using De-ess.


  • Algorithm:Classic reduces the volume of all frequencies on an Ess sound, while Spectral reduces the frequency of the band most associated with Ess sounds.
  • Threshold:This helps set the threshold point and will help reduce Ess sounds that go over the volume. Of course lowering it makes the effect more obvious. Though we can look at the gain reduction meter for reference, but by ticking the Output Ess only box here, we can hear just the sounds that are being detected. Adjust it to try and remove just the “Ess” sounds.
  • Cutoff Frequency:Determines the cutoff point for what is analysed as an Ess sound and all other sounds, and raising it narrows down the target frequencies, while lowering it applies the effect to include lower frequency Ess sounds. We can use Output Ess only in this instance as well.
  • Speed:Determines the attack/release of the De-esser, with a Fast or Slow setting. Though Fast will most likely work in most situations, we can use Slow if the effect doesn’t work well with fast. If Slow makes the reduction too obvious with a pumping effect, use Fast.
  • Spectral shaping:Can be considered a detailed adjuster for the strength of the De-esser effect. If the de-esser sounds too strong, lower this fader.
  • Spectral tilt:Determine the noise profile of the Ess sounds. Default at 0 is Pink noise which is a natural sound with more midrange. Pulling it to the left targets Brown noise which contains more low frequencies. Pulling it to the right targets White noise which contains more high frequencies.

This completes our noise removal edits.
Here’s how it sounds before and after editing and removing noise.



The sound has been substantially improved and cleaned up.

This completes our look at the most commonly used modules in RX7.
Try these out to bring your troubled recordings back to life!