Denoiser is designed to reduce stationary noise, including broadband noise, in audio signals. Stationary noise can include tape hiss, microphone hum, power mains buzz, camera noise and other types of noise that do not change in level or spectral shape throughout the recording.
Training Denoiser:
Denoiser can learn the type of noise you want to remove from the recording to give you the best results. To train Denoiser, select a section of the recording that contains only noise and click Train. This will create a noise profile that Denoiser will use to process the recording. If you do not have a section of the recording that contains only noise, selecting a quiet region of the recording that contains mostly noise can also work.
Note: The noise profile will be valid only for the currently selected FFT size,
current sampling rate and number of channels. If either of these parameters
changes, the noise profile should be re-collected with the updated parameters
for accurate noise reduction results.
Noise Spectrum Display:
Noise Spectrum display shows useful information during both playback and when the noise reduction process is being applied:
* Red curve - spectrum of input audio signal
* Yellow curve - spectrum of Denoised (output)
audio signal
* Blue curve - noise threshold, it equals learned noise profile + Threshold
elevation control
* Light blue curve - desired noise floor after denoising,
can be controlled by Reduction and Residual whitening controls (advanced mode only)
Scrolling and Zooming - by grabbing and moving the horizontal and vertical rulers, you can adjust what part of the graph is shown. In addition, holding the pointer above a ruler and using your mouse wheel will zoom the display in and out.
Simple Mode Controls:
Algorithm - affects the quality and computational complexity of the noise reduction. This selection directly affects CPU usage.
Noise Reduction (dB) - controls the desired amount of noise suppression in decibels.
Artifact Suppression - helps minimize "musical noise" artifacts introduced during the noise reduction process.
Residual output - outputs the difference between the original and processed signals (suppressed noise)
Advanced Mode Controls:
Threshold (tonal/broadband) - controls the separation of noise and useful signal levels.
The higher the threshold is set, the more noise is reduced. However setting threshold too high can also remove or suppress low level signals.
Higher threshold settings reduce more noise, but also suppress low-level signal components. Lower threshold preserves low-level signal details, but can result in noise being modulated by the signal. Threshold elevation can be done separately for tonal and random noise parts. A good default is 0 dB.
Reduction (tonal/broadband) - controls the desired amount of noise
suppression in decibels.
Denoiser can automatically separate noise into tonal parts (such as hum, buzz or interference) and random parts (such as hiss). The user can specify the amount of suppression for these parts separately (e.g. in some situations it can be desirable to reduce only unpleasant buzz while leaving unobjectionable constant hiss). Strong suppression of noise can also degrade low-levels signals, so it's recommended to apply only as much suppression as needed for reducing the noise to levels where it becomes less objectionable.
Musical noise suppression - controls the reduction of artifacts known as
"musical noise".
Musical noise is caused by random statistical variations of noise spectrum that cause random triggering of sub-band gates resulting in "metallic" or "space monkey" sounds. The control synchronizes triggering of noise gates in time and in frequency and makes gates sensitive only to larger sound events eliminating small random variations. Too high values of this slider can produce muffled sound and suppress minor details in the useful signal.
Residual Whitening - modifies the amount of noise reduction (shown by
light blue curve) applied at different frequencies to shape the spectrum of the
residual noise.
When residual whitening is zero, the suppression is uniform at all frequencies, as controlled by Reduction (tonal/broadband) sliders, and the suppressed noise has a similar spectral shape to the original noise. When residual whitening is maximal, the desired shape of suppressed noise floor is made close to white, so that residual noise has more neutral sound.
MNS algorithm - Musical noise artifact suppression algorithm
* Simple algorithm performs independent noise gating in every frequency channel of FFT. Release time of sub-band gates is controlled by the Release control. This is a fast algorithm with small latency that is suitable for real-time operation.
* Advanced and Extreme algorithms perform joint time-frequency analysis of the audio signal which results in better quality and less "musical noise" artifacts. These algorithms have higher latency and computational complexity.
FFT size (ms) - controls the FFT window size of the denoiser.
Longer windows mean lower time resolution, but better frequency resolution. Good frequency resolution (longer windows) allows for better separation of tonal signals and noise, but it also reduces time precision of Denoiser which can lead to softened transients and less noise reduction around transients. This control has no effect in Multi-resolution mode.
Whenever FFT size is changed, it's recommended that the user retrain Denoiser because the old noise profile was taken at a different FFT size and therefore becomes inaccurate.
Multi-resolution - enables multi-resolution versions of algorithms
(applicable to any selected algorithm type)
The multi-resolution processing selects the best time-frequency resolution for every portion of the signal to minimize smearing of transients and at the same time achieve high frequency resolution where it is needed. The overall quality of noise reduction is increased, but at the cost of much higher computational complexity.
The FFT size control does not have any affect in multi-resolution mode because FFT resolution is selected automatically. Noise profile does not need to be re-learned when switching a Multi-resolution checkbox.
Knee sharpness - adjusts the sharpness of the noise gates (1 is soft knee, 8 is sharp knee)
Release - selects the release time of sub-band noise gates in milliseconds
Longer release times reduce musical noise more effectively, but can produce echoes of noise after loud signals.
Psychoacoustic suppression - enables a psychoacoustic model that
dynamically controls suppression amount to facilitate the use of softer
suppression where noise is subjectively inaudible.
The maximal strength of suppression is still given by the light blue curve, so the psychoacoustic model can only dynamically reduce amount of suppression for more careful noise reduction and less signal distortion in certain regions. The position of the slider controls the influence of psychoacoustic model on suppression levels.
Harmonic enhancement - allows for softer suppression of harmonics of the
signal that could be buried in noise.
It can make the resulting signal brighter and more natural sounding, but high values of harmonic enhancement can result in high-frequency noise being modulated by the signal.