> ## Content Index
> Fetch the complete content index at: https://soundforgepro.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# RX Dialogue Isolate in RX 12: Controls, Settings and Test Results
- URL: https://soundforgepro.com/izotope-rx-dialogue-isolate/
- Published: 2026-10-20T22:30:16.000Z
- Updated: 2026-10-22T01:08:14.000Z
- Description: RX 12 Dialogue Isolate splits speech from noise and room reverb with three faders. What each control does, settings that keep the voice natural, Standard vs Advanced, plug-in latency, Stem Split and what published comparisons found.
- Author: Erick Finn
- Tags: iZotope RX, Voice Recording

**RX Dialogue Isolate is iZotope's machine-learning speech separator: it splits a recording into voice, reverb and noise, and you set the level of each part with three faders. In RX 12 it ships in RX Standard and RX Advanced, as an RX Audio Editor module and as a DAW plug-in; RX Elements does not include it.** Advanced adds a Quality menu with an offline mode and a four-band panel, while Standard runs without them. According to [iZotope's RX release notes](https://www.izotope.com/pages/release-notes/rx-advanced), RX 12.0.0 arrived on April 28, 2026, and the current 12.1.0 update from October 15, 2026 fixes an empty expanded advanced panel in the Dialogue Isolate plug-in after hide/show in Logic Pro running natively.

The routine I recommend barely changes from clip to clip: audition in real time, render the keeper offline, keep Voice at 0 dB, pull Noise and Reverb down separately and touch Sensitivity last. I would also treat Dialogue Isolate as a post-production tool, not a live insert: iZotope support says as much, and the published latency readings back it up. The sections below walk through each control, the Standard and Advanced differences, the published test results and the cases where another RX module is the better first move.

## What Dialogue Isolate does in RX 12

The [RX 12 Dialogue Isolate manual page](https://docs.izotope.com/rx12/en/dialogue-isolate.html) describes a model trained on a large library of speech, noise and reverberant recordings. It detects dialogue, noise and reverberation and separates them into three signal components, and the Voice, Reverb and Noise gain controls set how loud each component comes back. iZotope also notes that it can be particularly effective at raising the level of dialogue in recordings with a poor signal-to-noise ratio.

The same page lists what it is built for: constant or non-stationary background noise such as hiss, crowds, traffic, footsteps and weather, plus the echo and reverberation of spaces such as a tiled room or a lecture hall. That second part matters. You do not need a separate de-reverb pass on most dialogue, because the Reverb fader handles the room in the same module.

It is a dialogue tool first. For singing voice, instruments or full mixes, iZotope points to Spectral De-noise, which is not focused on speech. If you are planning a full cleanup chain, the [noise-removal workflow across RX modules](https://soundforgepro.com/remove-background-noise-izotope-rx/) covers the order of operations; this page stays inside Dialogue Isolate, one part of the wider [iZotope RX toolkit](https://soundforgepro.com/izotope-rx/).

## Standard vs Advanced inside the module

The edition split for this one module is small but it changes how you work. The [RX 12 overview and edition tables](https://docs.izotope.com/rx12/en/rx-overview.html) list Dialogue Isolate under Standard and Advanced only, note that Elements is six plug-ins without the RX Audio Editor, and state that the Advanced version adds a high-quality offline mode and multi-band processing.

| Feature                                         | RX 12 Standard | RX 12 Advanced | RX 12 Elements       |
| ----------------------------------------------- | -------------- | -------------- | -------------------- |
| Dialogue Isolate module in the RX Audio Editor  | Yes            | Yes            | No (no Audio Editor) |
| Dialogue Isolate plug-in                        | Yes            | Yes            | No                   |
| Voice, Reverb and Noise faders, Sensitivity     | Yes            | Yes            | No                   |
| Quality menu (Good / Real-time, Best / Offline) | No             | Yes            | No                   |
| Four-band multi-band panel                      | No             | Yes            | No                   |
| Stem Split and Stems view                       | Yes            | Yes            | No                   |
| Voice De-noise, the zero-latency alternative    | Yes            | Yes            | Plug-in only         |

[ProVideo Coalition's RX 12 shootout](https://www.provideocoalition.com/izotope-rx-12-shootout/) sums up Standard as including Dialogue Isolate "but only the lower quality mode", with the higher-quality offline mode reserved for Advanced. Some older marketing copy still frames Dialogue Isolate as an Advanced-only specialty; the manual is the reference to trust.

Which edition to buy depends on far more than this module, so I keep that decision in the [RX 12 Elements vs Standard vs Advanced comparison](https://soundforgepro.com/izotope-rx-elements-vs-standard-vs-advanced/). Current prices, upgrade paths and the trial live in the [RX price, trial and license guide](https://soundforgepro.com/izotope-rx-price-trial-license/).

## The controls, one by one

![RX 12 Dialogue Isolate module with Voice at 0.0 dB, Reverb at -14.0 dB and Noise at -12.8 dB on faders scaled from 20 to -Inf, a Sensitivity slider at 5.0 and footer buttons Preview, Bypass, plus, Compare, Stem Split and Render; orange markers 1 to 4 point to the Voice fader, the Sensitivity slider, Stem Split and Render](https://storage.ghost.io/c/af/4b/af4b05d4-2e0b-4edc-8378-1e2a62fecc70/content/images/2026/10/dialogue-isolate-controls.webp)

Leave Voice (1) at 0 dB, bring Reverb and Noise down separately, move Sensitivity (2) last, then split stems (3) or render (4).

Every control in the Standard module is on that one panel. Here is what each one does, using the wording of the RX 12 manual.

- **Voice.** Gain in dB applied to whatever the model identifies as voice. The manual's module view shows the three faders on a scale from +20 dB down to -Inf, so Voice can lift quiet speech as well as cut it.
- **Reverb.** Gain in dB for the reverberation component. Pulling it down dries the room out.
- **Noise.** Gain in dB for the noise component. This is the fader most people mean when they say "cleanup".
- **Sensitivity.** How much of the original audio counts as noise and reverb. Raising it treats more of the signal as noise and reverb and less as voice, which can give deeper reduction but may add artifacts and reduce dialogue clarity.

One caution about older tutorials. The [RX 9 Dialogue Isolate help page](https://s3.amazonaws.com/izotopedownloads/docs/rx9/en/dialogue-isolate/index.html) described Sensitivity from the dialogue side: higher values defined dialogue more broadly and let more noise through. The RX 12 manual describes the slider from the noise and reverb side. If an RX 9-era video tells you which way to move it, check the current manual's wording and your ears before copying the move.

The footer follows the [common RX module controls](https://docs.izotope.com/rx12/en/common-module-controls.html): Preview, Bypass, Compare (which opens the Compare Settings window for auditioning several settings) and Render, plus a Stem Split button that is specific to this module. In plug-ins, Bypass is the only one of those footer controls you keep, so the Compare Settings window is an Audio Editor convenience.

## Settings that keep the voice natural

There is no universal preset for dialogue. The right depth depends on how loud the noise is against the voice, so work in an order rather than toward a number.

1. Select a representative stretch: a few seconds that contain both speech and the noise you want gone.
2. Click Preview with Voice at 0 dB. Leave it there; Voice sets the speech level, not the cleanup.
3. Pull Noise down in moderate steps until the background sits where you want it. Stop as soon as consonants or breaths start to sound thin.
4. Do the same with Reverb, on its own. Move it more cautiously than Noise, because a fully dry voice can sound detached from a picture shot in a real room.
5. Move Sensitivity last, in small steps, and back off at the first hint of watery or chirpy artifacts.
6. In the Audio Editor, open Compare and audition two or three candidate settings against each other before you render.
7. If one pass is not enough, render a gentle pass and run a second one rather than forcing the first. [iZotope's dialogue-cleanup guide](https://www.izotope.com/community/blog/how-to-remove-background-noise-from-dialogue-recordings) makes the same point: a couple of gentle passes are more transparent than one extreme pass, and keeping a copy of the original helps you catch artifacts.

The values in the module image above, with Reverb near -14 dB and Noise near -13 dB, are a moderate example state, not a recipe. Treat your ears and the before-and-after comparison as the final judge.

## Good / Real-time vs Best / Offline

This menu exists only in RX Advanced. The manual defines the two modes clearly:

- **Good / Real-time** is a fast, high-quality mode optimized for low latency and plug-in use. The collapsed menu simply reads "Real-time".
- **Best / Offline** is iZotope's highest-quality mode, with fewer artifacts and better isolation, at the cost of longer processing and higher latency. It is meant for the RX Audio Editor and for offline bounces or renders in a DAW.

Use them as a pair: audition in Real-time, switch to Best / Offline, then render. [Production Expert's 2025 dialogue shootout](https://www.production-expert.com/production-expert-1/dialogue-noise-reduction-shootout-2025-the-results) handled the plug-in exactly that way, auditioning settings in Good / Real-time and rendering in Best / Offline. Since RX 11.1.0 (June 6, 2024), choosing Best / Offline inside a real-time plug-in shows a one-time warning that reminds you to use the real-time mode for playback.

## Four-band processing in Advanced

![RX 12 Advanced Dialogue Isolate Quality menu set to Real-time beside the multi-band display with a 12 to -24 dB scale, a frequency axis from 60 Hz to 10 kHz, three crossover lines with round handles and four knobs at 100 labeled Low, Low Mid, High Mid and High; marker 1 points to the Quality menu, markers 2 and 3 to crossover handles](https://storage.ghost.io/c/af/4b/af4b05d4-2e0b-4edc-8378-1e2a62fecc70/content/images/2026/10/dialogue-isolate-advanced-panel.webp)

Advanced only: set Quality (1) before rendering, and drag the crossover handles (2, 3) to decide where the four bands split.

The Expand button at the bottom right of the Advanced module opens metering and a multi-band panel. It scales processing from 0 to 100% in four bands, Low, Low Mid, High Mid and High, and the three crossover points are adjustable. The manual's own example: with Noise at -10 dB, a band set to 0% leaves the detected noise in that band at 0 dB, while 100% applies the full -10 dB. Difference metering shows how much gain change is happening in each band.

When any band sits below 100%, the Expand button turns blue. That is a useful reminder when you reopen a session and wonder why a preset behaves differently on a new clip.

The bands earn their place when the noise lives in one region. In iZotope's RX 11 cleanup example, wind noise sat mostly under 100 Hz and crickets around 3.2 kHz, and the engineer expanded the plug-in to focus processing on those two areas. The same panel helps when processing goes too far. If the voice starts to lose body, I would lower the Low band percentage first, and if sibilants turn dull or lispy, check High Mid and High before touching Sensitivity.

## Plug-in latency and live use

The plug-in works in DAWs and NLEs, but it is not a zero-latency insert. The [RX 12 plug-in guide](https://docs.izotope.com/rx12/en/using-rx-plug-ins.html) explains that some processing modes are CPU-intensive and delay the signal; hosts correct this with delay compensation, and RX displays the delay as Total System Delay so you can compensate manually if your host does not.

iZotope's own position is plain. In a [Native Instruments community thread on RX 12 Dialogue Isolate latency](https://community.native-instruments.com/discussion/50782/rx-12-dialogue-isolate-latency), a support representative wrote on May 12, 2026 that the plug-in is not recommended for live work where minimal latency is required, that it is best used offline in Best / Offline in the RX Editor, and that the host must compensate for its latency.

Published readings are scarce and each comes from a single source, so I report them as they were published rather than as a median:

- [Production Expert's RX 11 review](https://www.production-expert.com/production-expert-1/izotope-rx-11-review-deep-dive-from-post-expert) gave the real-time AAX plug-in a latency of 2048 samples, the same as Waves Clarity Vx and lower than dxRevive and Extract: Dialogue. The sample rate was not stated; run that figure through the [samples-to-milliseconds latency calculator](https://soundforgepro.com/audio-latency-calculator/) at your own session rate.
- In the same Native Instruments thread, one user reported that the Live Professor 2 host showed 42.3 ms in RX 11 and 85.3 ms in the RX 12 trial. Sample rate, buffer and Quality mode were not given.

For live monitoring, streaming or tracking, use [Voice De-noise](https://docs.izotope.com/rx12/en/voice-de-noise.html) instead. It is a zero-latency adaptive de-noiser and it is in all three RX 12 editions, as a plug-in only in Elements.

On the compatibility side, the RX 12.0.0 notes list AU, AAX, AAX AudioSuite and VST3 formats, all 64-bit only, on macOS Sonoma 14.7, Sequoia 15.7 or Tahoe 26.4 and on Windows 10 (22H2) or Windows 11 (24H2), with Intel and Apple silicon Macs supported. Listed hosts include Pro Tools 2025, Logic Pro 12, Ableton Live 12, Cubase 15, Nuendo 15, Reaper 7, FL Studio 25, Adobe Audition 2026, Premiere Pro 2026 and DaVinci Resolve 19\. If the plug-in does not appear in your host after installing, work through the [fixes for RX plug-ins not showing](https://soundforgepro.com/izotope-rx-plugins-not-showing/).

## Stem Split, Stems view and room tone

The Stem Split button in the module footer splits the selection into three stems, dialogue, noise and reverberation, and opens them in a dedicated Stems tab. There you can process each stem separately and export it. One rule catches people out: Stem Split ignores the Voice, Reverb and Noise fader positions and always assumes 0 dB, so you get clean components, not your balanced mix.

The [RX 12 Stem Split pane](https://docs.izotope.com/rx12/en/stem-split-pane.html) offers the same job from the main window through a Dialogue Isolate minimodule. You set Sensitivity and, where your edition offers it, Quality, then click Process. Stems can be exported one at a time, all at once as separate files, or combined with Export as Single File. Stems view itself is new in RX 12 and available in Standard and Advanced through Dialogue Isolate, Music Rebalance or Scene Rebalance, the last of which is an Advanced module.

The noise stem has a second life as fill. In the same iZotope guide, the engineer describes using presets that do the inverse job, removing dialogue and keeping the background, when a scene needs extra room tone. In RX 12 the direct route is the noise stem or Voice pulled to -Inf. If you edit in Sound Forge, the guide to [recording, editing and using room tone](https://soundforgepro.com/room-tone-sound-forge/) shows how to lay that fill under cuts.

## Dialogue Isolate or another RX module?

![Decision diagram starting from What is under the voice?: crowds, traffic, footsteps and weather lead to Dialogue Isolate; room reverb on speech leads to the Dialogue Isolate Reverb fader; steady hiss or buzz left over leads to Spectral De-noise; live monitoring with no added delay leads to Voice De-noise; reverb on music or effects leads to De-reverb](https://storage.ghost.io/c/af/4b/af4b05d4-2e0b-4edc-8378-1e2a62fecc70/content/images/2026/10/rx-module-decision-path.webp)

Start from what sits under the voice: Dialogue Isolate covers changing noise and room reverb, and the other modules cover the rest.

Dialogue Isolate is the right first move more often than not, but it is not the only RX answer. The manual itself suggests [Spectral De-noise](https://docs.izotope.com/rx12/en/spectral-de-noise.html) for stationary noise when Dialogue Isolate does not give acceptable results, and [De-reverb](https://docs.izotope.com/rx12/en/de-reverb.html) for general reverb that is not on dialogue.

| Problem under the voice                             | First module to try            | Why                                                                                                       |
| --------------------------------------------------- | ------------------------------ | --------------------------------------------------------------------------------------------------------- |
| Crowds, traffic, footsteps, weather                 | Dialogue Isolate               | Built for non-stationary noise that keeps changing                                                        |
| Room reverb on dialogue                             | Dialogue Isolate, Reverb fader | Reverb removal is part of the same model                                                                  |
| Steady hiss, buzz or line noise left over           | Spectral De-noise              | iZotope's own suggestion; also works on singing, instruments and mixes                                    |
| Reverb on music or effects                          | De-reverb                      | Dialogue Isolate is speech-specific                                                                       |
| One-off events such as creaks or camera focus noise | Spectral Repair                | Attenuates or replaces a selected event; won the camera-noise test in ProVideo Coalition's RX 11 shootout |
| Live monitoring, streaming, tracking                | Voice De-noise                 | Zero latency and included in every edition                                                                |

Echo across a whole recording is a slightly different job; the guide on [removing echo from audio without wrecking speech](https://soundforgepro.com/remove-echo-from-audio/) covers it across tools. If you edit in Sound Forge and do not own RX Standard, start with [Sound Forge's own noise reduction](https://soundforgepro.com/how-to-remove-background-noise-in-sound-forge-pro/) before buying anything; the [Sound Forge Pro vs iZotope RX comparison](https://soundforgepro.com/sound-forge-pro-vs-izotope-rx/) covers when the extra repair tools earn their price.

## Test results: how Dialogue Isolate fares in published comparisons

The composite result from four published comparisons by Production Expert and ProVideo Coalition, run between 2024 and 2026, is clear. Dialogue Isolate was the top pick in 2 of 15 scenarios, finished first in 0 of 4 traffic scenarios, and was praised for keeping the voice natural rather than for the deepest reduction. Every one of the four reviews that compared versions heard it improve, though they disagree on how much the RX 12 step adds.

| Comparison                        | Date            | RX version                | Scenarios | Dialogue Isolate result                                |
| --------------------------------- | --------------- | ------------------------- | --------- | ------------------------------------------------------ |
| Production Expert listening test  | January 2, 2024 | Not stated (before RX 11) | 4         | 0 wins; third of six tools on total votes              |
| ProVideo Coalition RX 11 shootout | May 24, 2024    | RX 11                     | 4         | 1 win, on keeping the voice's original tone            |
| Production Expert shootout 2025   | June 27, 2025   | RX 11 era (not stated)    | 4         | 0 wins; one second, two thirds and one fourth of eight |
| ProVideo Coalition RX 12 shootout | May 1, 2026     | RX 12                     | 3         | 1 win, on steady background noise                      |

In [Production Expert's dialogue restoration listening test](https://www.production-expert.com/production-expert-1/results-dialogue-restoration-listening-test-noise-reduction), readers voted blind on four clips processed by six dialogue noise-reduction tools. Dialogue Isolate drew 369 of 2,328 votes, third overall, without winning a single clip. On the hum clip the result was "practically a draw" between Extract: Dialogue, Dialogue Isolate and dxRevive Pro; on roadside traffic, Hush Pro won and Dialogue Isolate came third.

Production Expert's 2025 shootout widened the field to eight tools and four examples. Its summary gives Dialogue Isolate "second place with example 1, two third places, and one fourth place", and its verdict is that "RX didn't do badly, but no longer dominates". Hush Pro, dxRevive Pro and Supertone Clear took the wins, and the conclusion was "The goal is a toolbox".

The ProVideo Coalition tests explain why RX still holds its place. In the [RX 11 shootout](https://www.provideocoalition.com/izotope-rx-11-shootout/), with settings pushed hard on most tests and the multi-band panel unused, Dialogue Isolate was the favorite on the first test for "keeping the original tone of the voice", joint second behind Adobe Enhance Speech on the second test, beaten on heavy traffic plus reverb by Waves Clarity Vx, and beaten on camera focus noise by RX's own Spectral Repair. In the RX 12 shootout it won the steady background noise test because "RX is truer to the original", while Adobe Enhance Speech online won the phone-with-reverb and heavy-traffic tests. On heavy traffic, RX 12 had "significantly less artifacting" than RX 11, though an occasional booming reverb artifact still rang through.

On the version steps, opinions differ in size, not direction. In its [RX 11 review](https://www.production-expert.com/production-expert-1/izotope-rx-11-review-deep-dive-from-post-expert), Production Expert found the new module removed noise "notably more effectively" than before and handled reverb while "impacting the voice less" than the old Dialogue De-reverb. For RX 12, ProVideo Coalition hears a noticeable improvement over RX 11, "but granted it's not massive". [MusicTech's RX 12 review](https://musictech.com/reviews/plug-ins/izotope-rx-12-review/) is warmer, calling the result "markedly higher quality than previously - as convincing as any I've heard".

My verdict: Dialogue Isolate is the first tool I would reach for when the voice has to stay natural, and in RX 12 it is the best version of the module so far. It is not the strongest separator for heavy traffic, so on a hard traffic clip I would still start with it and then compare at least one other tool; the [best iZotope RX alternatives](https://soundforgepro.com/best-izotope-rx-alternatives/) page covers the candidates. Pair it with Spectral Repair for one-off events, and read the full [iZotope RX 12 review](https://soundforgepro.com/izotope-rx-12-review/) for how the rest of the suite holds up.

## When the result sounds watery or muffled

- **Watery, chirpy or phasey speech.** Sensitivity is too high or Noise is pulled too far. Both raise the risk of artifacts and reduced clarity, so back them off before anything else.
- **A voice that sounds pasted on.** Reverb is too low for the picture. Bring it back up a few dB until the voice sits in the room again.
- **The end of an offline selection sounds muffled.** That was a known RX 11 bug, fixed in RX 11.2.0 on September 20, 2024\. If you are on an older 11.x build, update.
- **A warning or stutter in a plug-in.** You have chosen Best / Offline in a real-time plug-in. Switch to Real-time for playback and keep Best / Offline for the render.
- **Dialogue out of sync with picture.** Check that your host's delay compensation is on and compare it with the Total System Delay RX reports.

## How the module changed from RX 6 to RX 12

Dialogue Isolate first appeared in RX 6 and was updated in RX 10\. Until RX 11, the plug-in was AudioSuite-only and the module belonged to the top edition. [Production Expert's RX 11 review](https://www.production-expert.com/production-expert-1/izotope-rx-11-review-deep-dive-from-post-expert) recounts that history.

RX 11.0.0, released May 14, 2024, was the big rebuild. The release notes list a new neural net in the RX application, the inclusion of Dialogue De-reverb, the high-quality offline mode, multi-band processing and a new real-time Dialogue Isolate plug-in, and RX 11 Standard gained the module for the first time. RX 11.1.0 added the Best / Offline warning in real-time plug-ins, and RX 11.2.0 fixed the muffled end of offline selections.

RX 12.0.0 followed on April 28, 2026\. ProVideo Coalition notes updated neural nets and new plug-in versions for Dialogue Isolate, and Stems view arrived in Standard and Advanced. The real-time plug-in and Standard availability are RX 11 features, not RX 12 news. Whether the RX 12 step is worth paying for is covered in [RX 12 vs RX 11](https://soundforgepro.com/izotope-rx-12-vs-rx-11/).

## FAQ

### Is Dialogue Isolate in RX 12 Standard?

Yes. Standard includes the module and the plug-in with the Voice, Reverb and Noise faders and Sensitivity. Advanced adds the Quality menu with Best / Offline and the four-band panel. Elements does not include Dialogue Isolate.

### Can I use Dialogue Isolate live?

It is not recommended. The plug-in adds latency that your host must compensate, and iZotope support advises using it offline. For live use, Voice De-noise is zero latency and comes with every RX 12 edition.

### What does Sensitivity do in RX 12?

It sets how much of the audio the model treats as noise and reverb. Higher values remove more but can add artifacts and reduce clarity, so move it last and in small steps.