Download iZotope RX for Mac

In the edit

Audio editors, sound designers, post-production engineers and anyone who has ever had to salvage a noisy recording — this one is for you. iZotope RX is the audio restoration suite that finds its way into film and television workflows, music studios, podcast edit bays, video game audio pipelines and sample library production. It runs as a standalone application or inside your DAW, and it uses machine learning to analyse what it hears and suggest fixes before you even reach for a fader.

Dialogue Isolate, Music Rebalance, and a rebuilt De-bleed module

• Dialogue Isolate, Music Rebalance and Voice De-noise for separating and cleaning up specific elements
• De-click, De-clip, De-crackle, De-ess, De-hum and De-reverb for the usual suspects in bad audio
• A rebuilt De-bleed module in version 12 that handles crosstalk between mics more naturally
• Trim Silence and Stems View for working through long files and isolated stems

Dialogue editors, location sound mixers, and sample library builders rely on RX

If you edit dialogue for a TV drama, clean up location sound for a documentary, mix a podcast with guests on remote mics, or build sample libraries that need to be pristine, RX is the toolkit you reach for. The interface is in English, it runs on both Intel and Apple Silicon Macs under macOS 14.7 Sonoma, and it does not pretend to be simple — but it does pretend to get the job done, and it delivers.

How RX 11 sits alongside the competition

RX is the heavyweight in audio restoration, but it is not the only player. Two other names come up often in post-production circles: Accusonus ERA Bundle and Acon Digital Extract Dialogue. Both have their strengths, but they operate in slightly different spaces.

The ERA Bundle is lighter on system resources and quicker to set up. Its one-knob approach works well for podcasters and video editors who need a fast de-noise or de-clip without diving into spectral editing. Where ERA shines is speed — you can process a file in seconds and move on. Where it falls short is precision. If you have a recording with multiple overlapping problems — hum, crackle, and a distant rumble — ERA’s single-slider controls can only do so much. RX gives you separate modules for each issue, plus the Spectral Editor to paint out specific artefacts by hand.

Acon Digital Extract Dialogue is a more direct competitor to RX’s Dialogue Isolate module. It does one thing — separating speech from background noise and music — and it does it well. Extract Dialogue uses a different algorithmic approach that some engineers prefer for music-heavy content because it leaves fewer musical artefacts behind. The trade-off is that it lacks the broader toolkit. You cannot de-click a vocal take, de-hum a guitar amp, or trim silence across a four-hour interview inside Extract Dialogue. RX bundles all of those into one suite, which is why it stays the default choice for facilities that handle every kind of audio problem.

When the Repair Assistant overcorrects dialogue and how to fix it

Even experienced users hit the occasional wall with RX. Here are a few patterns that come up often and how to work around them.

The Repair Assistant overcorrects dialogue. The machine learning is good, but it can be too aggressive on voice recordings, especially if the speaker has a breathy or sibilant delivery. Instead of accepting the suggested module settings, run the Assistant to identify the problem areas, then switch to manual control and dial back the strength. You keep the detection work without losing the natural character of the voice.

De-clip introduces a hollow sound. Heavy clipping restoration can leave the waveform looking clean while the audio sounds thin. The fix is to use De-clip in smaller passes. Clip the worst peaks first at a moderate reduction, render that, then run De-clip again on any remaining clipped sections. Two gentle passes sound better than one heavy one.

Music Rebalance bleeds dialogue into the music stem. This happens most with dense mixes where the vocal sits deep in the arrangement. Try lowering the separation strength and using the output faders to blend the dialogue stem with the original mix. You lose some isolation but gain a cleaner sound. For critical separation, route the dialogue stem through Dialogue Isolate afterwards to clean up the residual music.

Spectral De-noise leaves a watery shimmer on quiet sections. That artefact usually means the noise profile was taken from a section that contained too much room tone or background activity. Capture a shorter noise profile — two to three seconds of pure silence or steady background hum — and reduce the algorithm’s sensitivity. If the shimmer persists, use the Spectral Editor to paint it out manually on the affected frequencies.

Cleaning up wind rumble and handling rustle from start to finish

Field recordings from location sound often combine low-frequency wind rumble with mid-range handling noise from the microphone cable or boom pole. Here is a start-to-finish approach inside RX 11.

Open the file in the standalone application and start with the Spectral Editor. Zoom into the low end, roughly below 100 Hz, and look for the broad, dark smear that indicates wind rumble. Use the brush tool to select that area and apply a gentle attenuation — do not remove it completely, just reduce it until it sits below the dialogue. Wind rumble that is removed entirely can leave a hollow gap in the low end.

Next, run the De-wind module if you have it, or use De-hum set to a low-frequency profile. The goal here is to catch any residual rumble that the Spectral Editor missed. Keep the reduction moderate so the low end of the voice stays intact.

Handling noise lives higher, usually between 200 Hz and 500 Hz, and sounds like a dull thud or rattle. Run De-crackle with a low threshold and a short crackle duration. This catches the transient clicks that handling noise creates without flattening the voice. If the noise is continuous rather than transient, use Spectral De-noise with a profile taken from a section of pure handling noise.

Finally, run Voice De-noise with a light touch to clean up any broadband noise that the previous steps did not address. Listen to the result on headphones and solo the difference signal to check that you have not removed too much room tone — a fully clean recording can sound unnatural for a field scene.

Set your output format and sample rate before loading any audio

When you open RX for the first time, take ten minutes to set up your preferences before you load any audio. Configure the output format and sample rate to match your usual project settings. This saves you from rendering files at the wrong specification later.

Learn the keyboard shortcuts for the modules you will use most. The standalone application supports custom shortcuts, so assign the Spectral Editor, Repair Assistant, and Voice De-noise to keys you can hit without looking. Muscle memory makes a real difference when you are processing a stack of files.

Start every session with the Repair Assistant on a short sample. Even if you plan to work manually, the Assistant gives you a baseline — it shows you what RX thinks the problems are and suggests module settings. You can then decide to follow its lead or go your own way. Over time, you will learn to predict what the Assistant will recommend, which means you are building a mental map of the suite’s behaviour.

Do not process your entire audio file in one pass. Break it into sections by problem type — wind noise here, electrical hum there, a plosive on one word. Process each section with the appropriate module at the right strength. A segmented workflow sounds cleaner and gives you more control than a single global pass.

Keep the original file untouched and work on a copy. This is obvious advice, but it is worth repeating because RX makes it easy to render destructive edits. Use the Save As command to keep your source file safe, and only render the processed version when you are happy with the result.

Emma salvages a coastal village boom recording with RX

Emma edits sound for a documentary series that films in remote locations. The crew uses lavalier mics on subjects and a boom for ambient sound. On a recent shoot in a coastal village, the boom operator captured ten minutes of a fisherman talking about the changing tides — but the audio had a persistent low-frequency hum from a nearby generator and occasional clicks from salt spray hitting the mic capsule.

Emma opened the file in RX 11 and ran the Repair Assistant. It identified the hum and the clicks correctly. She accepted the De-hum module at a moderate setting and let De-click handle the salt spray. The result was clean enough for the rough cut, but she noticed a slight boxiness in the lower midrange. She switched to the Spectral Editor, selected the 150–300 Hz range, and reduced the gain by 2 dB. That opened up the voice without removing the natural room tone that made the location feel real.

The whole process took about twelve minutes. Emma then exported a 48 kHz WAV file, dropped it into her DAW, and matched it to the picture. The director did not ask about the generator hum once. That is the kind of win RX delivers — the viewer never knows there was a problem.

New this release

• Scene Rebalance — separates dialogue, music and effects in a mix
• Trim Silence — strips quiet sections from long audio files
• Stems View — apply the full toolkit to isolated stems
• Rebuilt De-bleed and Breath Control modules

Mac it needs

• macOS 14.7 Sonoma
• Intel or Apple Silicon