Izotope Rx 6 Spectral Repair

Overview

Izotope Rx 6 Spectral Repair Kit
Izotope Rx 6 Spectral Repair Kit
Rx 6 Supplement

Spectral Repair intelligently removes undesired sounds from a file with natural-sounding results.

This tool treats selections within the spectrogram/waveform display as corrupted audio that will be repaired using information from outside of the selection.

Select the noise you want to repair and Spectral Repair will reduce it to the level of the noise floor, replace it with audio from around the selection, or generate entirely new audio to fit the selection.

Attenuate

This mode removes sounds by comparing what’s inside a selection to what’s outside of it.

Attenuate reduces spectrogram magnitudes in the selected area to match magnitudes from the surrounding area, resulting in the removal of the sound without leaving an audible gap behind.

Representing the latest advances in iZotope's assistive audio technology, Repair Assistant is a game-changing intelligent repair tool that can detect noise, clipping, clicks, and more. Solve common audio issues faster than ever, simply by selecting the type of material (music, dialogue, other) and letting RX 7 Advanced analyze the audio. Without access to audio recorded by an on-set location sound engineer, Mouthon has relied on iZotope RX over the years to clean up audio yet still retain the natural sound of whatever environment Bourdain is in. In this interview, Mouthon shares how he uses RX 6 Advanced on the show.

Since RX was first released in 2008, iZotope has produced a series of updates to its powerful spectral-based audio editor, making it one of the most effective audio-repair systems available. RX6 is possibly the biggest update of the application to date, with a series of new modules, as well as a new tier to the application in the form of RX. IZotope RX 6 Standard Audio Software iZotope RX 6 Standard is the industry standard audio repair tool that's been used on countless albums, movies, and TV shows to restore damaged, noisy audio to pristine condition. RX 6 Standard is a comprehensive audio repair and noise reduction toolkit that includes everything from RX Elements and more. Visually identify and repair problems with the standalone RX Audio Editor application, or use the plug-ins in real time in your favorite DAW or NLE.

Attenuate does not resynthesize any audio. It modifies dissimilar audio in your selection to be more similar to the surrounding audio.

Attenuate is suitable for recordings with background noise or where noise is the essential part of music (drums, percussion) and should be accurately preserved. It’s also good when unwanted events are not obscuring the desired signal completely.

For example, Attenuate can be used to bring noises like door slams or chair squeaks down to a level where they are inaudible and blend into background noise.

Attenuate: Controls

BANDS: Selects the number of frequency bands used for interpolation.
- A higher number of bands can provide better frequency resolution, but also requires wider surrounding area to be analyzed for interpolation.
- A lower number of bands is ideal for processing short selections or transient signals.
SURROUNDING REGION LENGTH: Defines how much of the surrounding content will be used for interpolation.
STRENGTH: Adjusts strength of attenuation.
MULTI-RESOLUTION: The multi-resolution mode allows for better frequency resolution for the interpolation of low-frequency content and better time resolution for the interpolation of high-frequency content.
BEFORE/AFTER WEIGHTING: Gives more weight to the surrounding audio before or after the selection.
DIRECTION OF INTERPOLATION: Determines where the material used in the repair process is located in relation to the current selection.
- HORIZONTAL: Signal to the left and right of the current selection will be used for interpolation.
- VERTICAL: Signal above and below the current selection will be used for interpolation.
- 2D: Signal above, below, to the left and to the right of the current selection will be used for interpolation.
  Note: Direction Of Interpolation control availability
  - Replace, Pattern and Partials + Noise tabs only utilize Horizontal mode and do not display this option.

Replace

The Replace tab can be used to replace badly damaged sections (such as gaps) in tonal audio. How to make a loop in garageband ipad. It completely replaces the selected content with audio interpolated from the surrounding data.

Replace: Controls

BANDS: Selects the number of frequency bands used for interpolation.
- A higher number of bands can provide better frequency resolution, but also requires wider surrounding area to be analyzed for interpolation.
- A lower number of bands is ideal for processing short selections or transient signals.
SURROUNDING REGION LENGTH: Defines how much of the surrounding content will be used for interpolation.
MULTI-RESOLUTION: The multi-resolution mode allows for better frequency resolution for the interpolation of low-frequency content and better time resolution for the interpolation of high-frequency content.
BEFORE/AFTER WEIGHTING: Gives more weight to the surrounding audio before or after the selection.

Pattern

This mode finds the most similar portion of the surrounding audio and uses this to replace the corrupted audio.
Pattern mode is suitable for badly damaged audio with background noise or for audio with repeating parts.

Pattern: Controls

BANDS: Selects the number of frequency bands used for interpolation.
- A higher number of bands can provide better frequency resolution, but also requires wider surrounding area to be analyzed for interpolation.
- A lower number of bands is ideal for processing short selections or transient signals.
SURROUNDING REGION LENGTH: Defines how much of the surrounding content will be used for interpolation.
MULTI-RESOLUTION: The multi-resolution mode allows for better frequency resolution for the interpolation of low-frequency content and better time resolution for the interpolation of high-frequency content.
PATTERN SEARCH RANGE: Selects the length of the audio segment used in a search for a suitable replacement interval. For example, setting it to 5 seconds will allow search within ±5 second range from the selection.

Partials + Noise

The advanced version of Replace mode. It restores harmonics of the audio more accurately with control over the Harmonic sensitivity parameter.

This mode allows for higher-quality interpolation by explicit location of signal harmonics from the 2 sides of the corrupted interval and linking them together by synthesis.

Partials + Noise is able to correctly interpolate cases of pitch modulation, including vibrato. The remaining of non-harmonic material (“residual”) is interpolated using Replace method.

Partials + Noise: Controls

BANDS: Selects the number of frequency bands used for interpolation.
- A higher number of bands can provide better frequency resolution, but also requires wider surrounding area to be analyzed for interpolation.
- A lower number of bands is ideal for processing short selections or transient signals.
SURROUNDING REGION LENGTH: Defines how much of the surrounding content will be used for interpolation.
HARMONIC SENSITIVITY: Adjusts amount of detected and linked harmonics.
- Lower values will detect fewer harmonics
- Higher values will detect more harmonics and can introduce some unnatural pitch modulations in the interpolated result.
MULTI-RESOLUTION: The multi-resolution mode allows for better frequency resolution for the interpolation of low-frequency content and better time resolution for the interpolation of high-frequency content.
BEFORE/AFTER WEIGHTING: Gives more weight to the surrounding audio before or after the selection.

Surrounding Region Display

SURROUNDING REGION SHADING: When using the Spectral Repair module, your selections will be shown with a dotted line surrounding your selected region. This dotted line is directly controlled by the Surrounding Region and Before/After Weighting controls inside of your Spectral Repair modules, and provides a visual representation of your set values.
- The surrounding region is the region that RX uses for interpolation of the selected region. The data from the surrounding region is used to restore the selected region.

Workflow

Applying Spectral Repair

To start working with Spectral Repair, switch to the spectrogram view by dragging the spectrogram/waveform transparency balance slider to the right.
Next, identify the unwanted event on a spectrogram and select it using a selection tool (use the time-frequency, brush, lasso or magic wand tools to fit your selection to the corrupted audio)
You can audition just the signal in your selection by pressing the Play Selection button in the RX transport.
Once you’ve found the event(s) to repair, select the appropriate Spectral Repair mode from the tabs at the top of the Spectral Repair settings window.
You can use the Compare Settings functionality to audition the processing before applying it, or click the process to apply the active Spectral Repair tab’s settings.

More Information

This section contains useful information, examples and tips for getting the most out of the Spectral Repair module.

Visual Example

The following image shows the selection before processing on the top and the selection after processing with Spectral Repair on the bottom.

Processing Limitations

Depending on the mode and settings, Spectral Repair will have varying limits to the amount of audio that can be processed in your selection.

Unlimited — Attenuate when in Vertical mode only.
10 seconds — Attenuate Horizontal or 2D; Replace modes.
4 seconds — Pattern, Partials + Noise modes.
Longer selections will automatically adjust processing to use the correct mode.

Spectral Repair as an alternative to De-click processing

When used with a time selection, Spectral Repair is able to provide higher quality processing than De-click for long corrupted segments of audio (above 10 ms).

Example use cases for Spectral Repair

When used with a time/frequency, lasso, brush, or wand selection, it can be used to remove (or attenuate) unwanted sounds from recordings, such as: squeaky chairs, coughs, wheezes, burps, whistles, dropped objects, mic stand bumps, clattering dishes, mobile phones ringing, metronomes, click tracks, door slams, sniffles, laughter, background chitchat, digital artifacts from bad hardware, dropouts from broken audio cables, rustle sounds from microphone movements, fret and string noise from guitars, ringing tones from rooms or drum kits, squeaky wheels, dog barks, jingling change, or just about anything else you could imagine.

Spectral Repair can also effectively repair gaps or replace audio intelligently by using advanced resynthesis techniques.

Increase efficiency with the Find Similar Event Tool

Some unwanted events consist of several separate regions on a spectrogram.
In some cases, it’s possible to achieve more accurate results by repairing several smaller selections one by one, instead of one large selection.
You can use the Find Similar Event tool to save time when searching for and fixing many similar events in large files.

Use the Compare Settings window to experiment with Spectral Repair processing

Sometimes it’s worth trying several different methods or number of bands to achieve the desired result. A higher number of bands doesn’t necessarily mean higher quality! We encourage you to use the Compare Settings window to experiment and find the best settings for the project at hand.

Adjust Surrounding Region Length and Before/after weighting to perfect your processing

Common parameters for many modes include Surrounding region length which determines how far around the selection Spectral Repair will look for a good signal. Before/after weighting allows you to use more information from either before or after the selection for interpolation. For example, if your unwanted event is just before a transient (such as a drum hit) in the audio, you may want to set this parameter to use more of the audio before the selection to prevent smearing of the transient into the selection.

iZotope's RX is one of the most powerful audio repair tools in the business. So what could a new version add? Quite a lot, as Joe Albano found out..

iZotope’s RX has established itself as the premiere audio repair software - its collection of cutting-edge tools and spectral processing capabilities has been used to clean up and improve countless bits of problem audio over the years. The latest version, RX6, adds a number of significant new features and processors to the party so let’s take a look at what’s on hand in this update.

The RX6 Family

First off, RX6 is now available in four versions. The top-of-the-line is RX6 Advanced, which contains all the processing modules available for RX, including the new ones, for the most full-featured set of tools. RX6 Standard offers a slightly smaller subset of the full toolbox for a reduced price, but it does contain most of the key modules, including the ones RX is best known for like Spectral Repair. For those on an even more modest budget, RX6 Elements is a smaller collection of basic repair tools, perfect for quickly cleaning up the most common flaws in typical recordings. A full comparison chart is available on the iZotope website. In addition to these options, for post production pros, RX6 Advanced is also available as part of the RX Post Production Suite 2, which also includes Neutron Advanced, Loudness Control and Insight.

General Features

RX6 (I’ll be describing RX6 Advanced here), just like previous versions of RX, consists of a standalone application with a collection of processing “modules”, each of which offers a targeted solution to a specific audio issue. Many of these are also available as plug-ins in all the standard formats (AAX, AU, VST), for use in realtime within your DAW of choice. When processing needs to be in the standalone (some of the more advanced types of audio tweakery, including the graphic techniques of Spectral Repair), RX6 provides a way to quickly transfer audio back and forth between the DAW and RX6 standalone - this varies in different DAWs, from using the RX Connect plug-in to assigning RX6 as the designated external audio editor.

Within the standalone, there are numerous features for enhancing workflow. Module Chains (savable as presets, as are all module settings) let the user assemble and quickly apply a string of modules, speeding up tasks that involve multiple processing. RX6 offers additional workflow improvements: an improved Find All Similar helps to quickly locate similar problems in a longer audio file, and there are new options for displaying the extensive module list by user category.

RX6 also introduces Composite View, a quick way to simultaneously process multiple files. While RX has always had batch processing capabilities via a dedicated window, Composite View is a more efficient way to temporarily group all open files together and apply processing to them all at once. While there are a few common sense restrictions, it could speed up workflow for many tasks that need to be performed across multiple bits of audio.

New Modules

Naturally, RX6 includes all the modules from previous versions, including (among others) De-click, De-clip, De-hum, De-plosive, De-reverb, Spectral and Voice De-noise, Deconstruct, and of course Spectral Repair as well as its collection of utility modules like Leveler, Loudness, Time & Pitch, Resample, EQ Match and many additional general-purpose modules. Some of these have incremental improvements - De-click, De-plosive, Voice De-noise, Ambience Match, Deconstruct and Center Extract all benefit from tweaks to their engines (I particularly noticed the difference in De-click).

To this already comprehensive collection of audio tools RX6 has added several new Modules: De-ess and Breath Control are broken out of the Leveler modules into separate modules with more comprehensive controls; Mouth De-click does the regular De-click one better, being targeted specifically to lipsmacks and other mouth-related clicks and noises; De-rustle and De-wind provide solutions for particular post production issues; Dialog Isolate pulls a voice up and/or out of a busy background; and De-bleed addresses that all-too-familiar problem of leakage, like from headphones worn during a recording. Each of these offers something specifically targeted to a particular problem so let’s take a quick look at them.

De-ess & Breath Control

In previous versions of RX, the Leveler module included simple ess-reduction and breath control options - basically just on-off and a sensitivity slider - but these new dedicated modules offer a lot more user control. De-ess provides settings for fine-tuning the frequency and speed of the de-essing, and it also offers a new Spectral De-essing mode, which applies spectral analysis and processing to the task of isolating and eliminating harsh sibilance. This can potentially make for more effective de-essing, with controls that allow the user to shape and tilt the spectral character of the sibilance, for more natural results. The new Breath Control module adds options for both amount of breath reduction and sensitivity to breath sounds, which seemed to me to isolate breaths even more precisely than the older version.

Mouth De-click

I’d never been able to get the regular De-click module in previous RX versions to reliably eliminate that problem on my own voice, but not only does the improved De-click do a better job, the dedicated Mouth De-click module works like a charm! Even at its default settings, this module seems to nail every bit of audio I’ve thrown at it, and it also provides additional controls to fine tune the detection and processing - whatever iZotope has done to tweak the general de-clicking algorithm to address dedicated mouth sounds, it seems to make a significant difference.

De-rustle & De-wind

These two are primarily for post production use, though of course they could come in handy on any application that suffers from those issues. De-rustle is aimed at removing or reducing the rustling sound of clothing rubbing up a lavalier mic. Like a number of the new modules, this addresses background noise that’s more irregular, and wouldn’t be handled as well by the standard De-noise modules, which do better learning & processing more steady, regular unwanted background sound. With a couple of files I threw at it, De-rustle was able to deal successfully with even fairly intrusive noise. A nice feature is an Ambience Preservation control, which allows De-rustle to remove rustling sounds while leaving other (possibly desirable) ambient sound intact.

De-wind addresses the problem of wind noise in outdoor recordings - the irregular and intrusive rumbling from wind shaking a mic’s diaphragm. With a couple of torture-test files I threw at it that suffered from especially loud wind noise, De-wind acquitted itself admirably - it should have no trouble with more typical subtle wind problems. A Fundamental Recovery control helps to restore any fullness in a voice lost to more aggressive processing, and Artifact Smoothing (a control common to many modules) helps to mitigate any hint of the watery-sounding artifact that sometimes accompanies this kind of FFT-based audio processing.

Dialog Isolate

This ambitious module lets you isolate a voice from background sound in a recording, or even vice-versa - pull a voice out of a recording, leaving just background sound (or music). Once again, unlike the existing Voice De-noise Module, Dialog isolate is tuned to be more effective with irregular background sound, rather than the regular noise the De-noise modules are designed to address.

And it works surprisingly well. I was able to substantially reduce the level of some especially intrusive background noise (subway train going by, loud voices is an ambient room), drawing out the primary voice until the background was, if not completely gone, largely unnoticeable; with less egregious examples of background sound, Dialog Isolate should be able to fully eliminate the undesirable noise. I wouldn’t necessarily expect it to be able to, say, completely lift a vocal out of a busy mix, especially one that’s somewhat buried in the mix, but considering how well it did isolating two voices (interviewer and subject) from background voices that were practically as loud as the main voices, it’s well worth trying on any recording - as with many of these processes, sometime multiple passes can provide an extra degree of effectiveness.

De-bleed

Finally, the new De-bleed module addresses the familiar issue of leakage, or bleed. As everyone knows, this can come from headphones during a recording (especially if a singer removes one earcup unnoticed, as they’re often inclined to do), in the form of the rough mix that was playing through the cans, or as the dreaded sound of a click track. Acoustic leakage, from other instruments in a live session, can also compromise the integrity of a track. But De-bleed can come to the rescue, if you’re careful to set it up properly.

De-bleed requires two files to do its thing: an “Active” track - the recording with the unwanted bleed - and a “Bleed Source” track - the track that contains the original audio that leaked into the Active track. For example, in a scenario with, say, an acoustic guitar recording that was cut to a click track in headphones, the guitar, naturally would be the Active track, and the original click track the Bleed Source. With a recording of a singer with bleed from a rough mix, the mix itself would be the Bleed Source. Both files must be loaded into RX6 (currently, De-bleed only works in the standalone version), and to be most effective, they need to be time-aligned. When I tested this, I bounced down the appropriate section of a leaked mix and lined it up with the vocal (which was an overdub) in the DAW, and then brought both files into the RX6 Advanced application.

And it worked! De-bleed Learned the profile of both tracks, and even though the bleed in my test file was quite loud (the singer had removed one earcup), and my Bleed Source mix file was not really the exact same rough mix as the one that had leaked into the vocal recording (I’d reconstructed it after-the-fact), De-bleed almost completely eliminated the leakage, especially under the vocal, where it really counted.

Wrap-up

So that’s what’s new and exciting in RX6 - while some of the new modules might seem at first glance to be merely refinements of existing features, in practice most offer significantly improved performance, along with much greater flexibility. And Dialog Isolate and De-bleed really do take audio repair to the next level. RX6 should be part of any audio engineer’s collection of tools, and with the various versions available, there’s no excuse not to have at least some of its features available for cleaning up your tracks and making them shine!

Price: RX6 Advanced: $1199 (regular); $799 (sale price)

RX6 Standard: $399 (regular); $299 (sale price)

RX6 Elements: $129 (regular); $99 (sale price)

RX Post Production Suite 2 (incl RX6 Adv): $1499 (regular); $999 (sale price)