Dialogue Isolate
STD & ADV | Module & Plug-in
Overview
Dialogue Isolate is designed to separate spoken dialogue from a constant or non-stationary background noise, such as hiss, crowds, traffic, footsteps, weather, or other noise with highly variable characteristics. It can be particularly effective at increasing the level of dialogue in challenging low signal-to-noise ratio recordings.
New in RX 11, Dialogue Isolate now includes reverb removal technology, which separates dialogue from the characteristic echo and reverberation generated by different spaces, such as a tiled room, a lecture hall, or a cathedral.
Dialogue Isolate uses a machine learning algorithm trained on a large library of speech, noise, and reverberant data. It automatically detects and separates dialogue, noise, and reverberation into three distinct signal components. The levels of the separated components can be independently adjusted using the Voice, Reverb, and Noise gain controls.
Controls (STD and ADV)
![]() |
Voice
Adjusts the amount of gain (in dB) applied to the components identified as voice.
Reverb
Adjusts the amount of gain (in dB) applied to the components identified as reverb.
Noise
Adjusts the amount of gain (in dB) applied to the components identified as noise.
Sensitivity
Determines how much of the noise and reverb will be extracted from dialogue by the separation algorithm.
Higher values will instruct the separation algorithm to broadly define what it categorizes as noise and reverb in the input signal. This can result in more significant noise reduction, at the cost of introducing artifacts and potentially reducing dialogue clarity.
Lower values will instruct the separation algorithm to narrowly define what it categorizes as noise and reverb in the input signal. This can result in more noise being included in the rendered signal, but can help to maintain dialogue clarity.
Quality
![]() |
The following quality modes are available in the Dialogue Isolate module.
Good / Real-time: A fast, high-quality processing mode that is optimized for speed (low-latency) and for use in the plugin.
Best / Offline: Our highest-quality processing mode with fewer artifacts and better isolation at the cost of a longer processing time (higher latency). For use in the RX Audio Editor and for offline bounces/renders in DAWs. (Advanced only)
Note
The RX 10 Dialogue Isolate legacy processing modes do not exist in RX 11.
Controls (ADV only)
The following controls are only available in the Advanced version of the Dialogue Isolate module and plugin (ADV):
![]() |
An expand button shows or hides the metering and multi-band interface by making the window larger or smaller.
If the reduction of any band is set to a value other than 100%, the expand button will be highlighted in blue to indicate that the multi-band controls will have an effect on the output.
Multi-band Interface
The multi-band interface in the Advanced module and the plugin adjusts the amount of processing across four different frequency bands (Low, Low Mid, High Mid, High). Each band is individually set from 0 to 100% of the processing parameter. (For example, if the Noise gain slider is set to -10dB, and a band is set to 0%, the gain of the detected noise in that band will be 0dB. If the band is set to 100%, the gain of the detected noise in that band will be -10dB.) The three crossover points are adjustable.
Difference metering provides visual feedback as to how much gain adjustment is happening across the four bands.
Alternative Modules
For stationary noise, such as hiss, buzz, line noise, etc., Dialogue Isolate should produce good results, but we also suggest trying the Spectral De-noise module if you are having trouble achieving acceptable results with Dialogue Isolate. Spectral De-noise is not focused on dialogue and can be used for a variety of source signals, like singing voice, instrumental tracks, or full mixes.
For more general, non-dialogue specific reverb reduction, try using the De-reverb module.