iZotope RX 12 Tutorial for Beginners: First Cleanup

Start learning iZotope RX 12 with a short problem recording: repair one audible defect and keep a route back to the original. That first pass teaches the part of RX that matters: hear the problem, locate it, select it, preview a treatment, compare, then render only when the result is better to your ears.

This tutorial uses the standalone RX Audio Editor in Standard or Advanced. RX Elements is a six-plug-in collection and does not include this editor. Seeing only plug-ins in a DAW does not identify your edition: Standard and Advanced also include plug-ins. Check the edition you installed before following the editor steps below. The official RX 12 feature comparison confirms that distinction.

What you need for this first RX cleanup

For this exercise, choose a 10–30 second mono or stereo WAV or AIFF file with one obvious problem. A spoken phrase with a click, steady fan noise, a short bump, or a clipped word is easier to judge than an entire song with five overlapping issues. Duplicate the source before opening it. RX keeps an edit history, but a separate untouched file protects you from a bad save or a mistaken export.

Open the copy with File > Open, Command+O on Mac or Ctrl+O on Windows, or drag it into the editor. RX supports common audio formats including WAV, AIFF, FLAC and MP3, although a lossless source is the better practice when you have one. The working-with-files section of the RX 12 manual lists the supported formats and platform caveats.

Before processing, play the clip from start to finish once. Write down the single defect that distracts you most. “The room is noisy” is still too broad; “a steady fan masks the ends of words” is useful. That sentence becomes your test: if the repair exposes the words without making speech watery, it worked.

Know which RX window you opened

RX Standard and Advanced include the standalone Audio Editor: a large waveform/spectrogram display, selection tools, modules, Preview, Compare and History. Elements supplies repair plug-ins for a DAW but no standalone editor. A beginner tutorial that shows spectral painting, a History pane or time-frequency selections therefore assumes Standard or Advanced.

Elements users can practice inside a host: loop a short passage, change one control, bypass the plug-in, and keep the setting that improves the defect while preserving the performance. The editor buttons and History steps below apply to Standard and Advanced. For a detailed purchase decision, use our RX 12 review and edition breakdown.

Read the waveform and spectrogram together

The waveform is strongest at showing level over time. The spectrogram adds frequency: time moves left to right, low frequencies sit at the bottom, high frequencies at the top, and the color scale represents amplitude. In the default palette, brighter orange and yellow indicate stronger energy; a custom color map can look different. RX overlays these views in one configurable display, as documented in iZotope’s spectrogram and waveform guide.

Use shapes as clues, not verdicts. A brief click often appears as a thin vertical streak because it spans many frequencies over very little time. Electrical hum often produces horizontal bands at a fundamental frequency and its harmonics. A steady noise floor can resemble a fine carpet behind the wanted signal. Repeated flat waveform peaks suggest clipping, but distortion has other causes, so listen and inspect before opening De-clip. The manual’s audio-problem examples explain these visual clues.

Move the waveform/spectrogram balance until both views answer a question. If the waveform makes the loud event obvious, lean toward it. If the noise occupies a narrow frequency band, show more spectrogram. You are not looking for a beautiful picture; you are finding the smallest region that contains the defect.

Make the smallest useful selection

Use the Time selection tool (T) for a time range across all frequencies, the Frequency selection tool (F) for a frequency band across time, or the Time-Frequency tool (R) for a rectangular region bounded in both dimensions. Keep Instant Process off while practicing: when enabled, it processes a new selection immediately with its chosen module settings. The RX selection-tool reference also documents modifiers for adding to, subtracting from and moving around a selection.

• Single click: start with a short time selection containing the click and nearby clean audio for De-click. For an isolated bump you intend to treat with Spectral Repair, a time-frequency box can limit the repair to its visible footprint.
• Steady fan or hiss: when using a manually learned noise profile, select background noise without speech or music you want to keep. After learning, switch to a test passage containing wanted audio. Adaptive modes analyze incoming audio instead.
• One clipped word: select the affected phrase plus a short lead-in and tail so Preview is meaningful.
• Whole-file noise: test on a difficult 3–10 second section before committing to the entire recording.

Beginners often select the whole file because it feels decisive. It can also apply an unsuitable setting far beyond the local defect. Prove the setting on a short, revealing passage first.

Choose one module from the sound you hear

Match the first module to a specific symptom. De-click is a sensible first test for short digital clicks or pops. De-clip attempts to reconstruct clipped peaks; it cannot promise to restore every distorted recording. Voice De-noise is a starting point for speech noise. Spectral De-noise offers a learned profile for stable noise and an Adaptive mode for changing noise. Spectral Repair uses surrounding audio to attenuate or replace an isolated event you can select; it can also alter wanted sound inside that region.

Do not stack modules merely because their names sound helpful. Every processor can suppress wanted detail along with the defect. One clear before/after comparison teaches you more than a chain whose individual effects you cannot identify. For a full noise decision tree, follow our RX 12 background-noise workflow; for visibly flattened peaks, use the De-clip guide.

Preview, bypass and Compare before you render

Open the chosen module and begin with a conservative preset or default. Preview applies to the active selection; with no selection, it starts at the playhead. During Preview, use Bypass to alternate between processed and original sound. Match your attention to the goal sentence you wrote earlier: are the word endings clearer, and did consonants stay intact?

Where the module provides a Listen or audition control, hear what the processor intends to remove. If that removed signal contains wanted syllables, drum attacks or musical notes, back off and check the result again. Turn Listen or the audition control off before judging the processed output against Bypass. Not every module supports Preview or Listen, so use Compare where available when direct Preview is unavailable.

Use Compare to store two settings on the same selection: one restrained, one stronger. Click Compare for the first, return to the module, change one control, and click Compare again. Select each entry in the comparison window and use its Preview button to hear that version. The official Preview and Compare documentation explains that Compare sends the current settings to its comparison window. Include the unprocessed original in your decision, and avoid mistaking “quieter” for “cleaner.” If one version improves the recording, select it and click Render in the comparison window. Keeping the original is also a valid outcome. Render applies processing; it does not export a delivery file.

Try one first repair: steady noise behind speech

This is a practice route for a noisy spoken phrase, not a preset that fits every recording. Select a short passage containing speech and a gap. Open Voice De-noise, enable Adaptive mode, choose Optimize for Dialogue, and start with the Gentle filter. Adaptive adjusts its threshold from incoming audio, so this route does not need a separate noise-only selection.

Preview the passage and increase Reduction only enough to make the noise less distracting. If consonants or quiet word endings lose their shape, reduce processing and compare with Bypass. Use Compare for a second setting if you are unsure. Render your chosen result once on the selected passage and listen across its boundaries.

If you want to apply the approved setting to the whole clip and the noise conditions are consistent, first return to the state before that test render in History. Select the whole clip with the Time tool, apply the chosen setting once, and listen from beginning to end. This avoids processing the test section twice. Different noise conditions need separate judgments; the linked background-noise guide covers those branches.

Use Repair Assistant as a starting suggestion

If you cannot name the problem yet, Repair Assistant is a useful second route. Choose Voice for spoken word or dialogue and Music for a mix, instrument or sung vocal, then use a selection containing representative speech or music. Click Learn on that selection. Assistant does not need a noise-only passage. That differs from manually learning a noise profile in Spectral De-noise; Spectral De-noise also has an Adaptive mode. iZotope’s Repair Assistant manual explains how it builds and exposes its suggestions.

Treat the result as a hypothesis. Bypass each active section and ask whether disabling it restores natural detail without returning the original distraction. Keep only the stages that earn their place. In RX 12 Advanced, Assistant recommendations can be opened as a Module Chain for deeper editing; that particular route is not available in Standard.

Keep the repair reversible with History and RX documents

After rendering, the History pane adds the edit to a vertical timeline. Click an earlier state to return to it and listen. To restore just a region, make the selection, hover over the desired earlier History entry, and click its Restore Selection icon. The rest of the file stays at its current state. That makes experimentation safer, but History is session state, not a substitute for an original backup.

If you need to resume later with History intact, choose File > Save RX Document… to create an .rxdoc file. The Undo History documentation states that RX documents retain the edit history. An RX document opens in the RX Audio Editor; export audio for a DAW or media player.

Choose File > Export for the entire file or File > Export Selection for only the selected region. Choose the format your next editor needs and a new name such as interview_fan-reduced_v1.wav. For this practice file, WAV keeps the handoff simple. Reopen the export and check its duration and beginning, middle and end: exporting only a selection by accident can leave you with a short excerpt.

Be careful with plain Save: for an opened WAV or AIFF it overwrites that file on disk. The manual’s save and export reference distinguishes this from Save RX Document. Keep the untouched source outside your working copy, even when you also keep an RX document.

Common first-cleanup mistakes

• Chasing silence: a little stable room tone is usually less distracting than metallic or watery speech.
• Trusting the picture alone: a bright mark can be wanted sound. Select with the display, decide with your ears.
• Judging only in a silent gap: noise reduction must also preserve consonants, breaths and the tail of words.
• Changing several controls at once: you lose the ability to identify which move helped.
• Skipping level-aware comparison: the louder version often feels better even when it is not cleaner.
• Rendering a full chain immediately: test one problem and one short passage first.

Repeat the first-cleanup exercise

1. Duplicate a short spoken-word file and open the copy in RX.
2. Play it once and describe the worst defect in one sentence.
3. Adjust the waveform/spectrogram balance until you can point to that defect.
4. Select the smallest useful test passage.
5. Open one matching module and preview a restrained setting.
6. Bypass it, then create a second, stronger version in Compare.
7. Listen for both improvement and collateral damage.
8. Render the chosen result once, inspect History, save an RX document if you need to resume, and export a newly named audio file.
9. Reopen the export and check that its duration and sound match the work you intended to deliver.

Repeat the exercise with a different defect on another practice file rather than adding five processors to this one. Once selection, Preview and comparison feel automatic, Module Chain and Instant Process become useful accelerators instead of shortcuts you cannot audit. The Learn iZotope RX hub organizes the next lessons by skill.

Frequently asked questions

Does RX Elements include the standalone Audio Editor?

No. RX Elements is a six-plug-in collection for a DAW. The standalone RX Audio Editor is included with RX 12 Standard and Advanced.

What should a beginner repair first in iZotope RX?

Choose the single defect that most distracts from the recording, then test one matching module on a short selection. A click, steady noise floor or clearly clipped phrase is easier to evaluate than a complex full-file chain.

Should I select the whole file before processing?

Only after a short test proves the setting. Isolated clicks, bumps and damaged phrases are safer to repair in focused selections; consistent whole-file noise can be processed globally after you check a revealing passage.

What is the difference between Preview and Compare in RX?

Preview lets you hear many module settings before rendering, usually on the active selection. Compare stores and auditions alternative settings or rendered previews, which is useful when direct Preview is unavailable or you need to judge several versions.

Should a beginner use Repair Assistant?

Yes, as a starting suggestion. Choose Voice or Music, let it analyze representative audio, then bypass each section and keep only the processing that fixes an audible problem without removing wanted detail.

How do I undo a bad RX repair?

Click an earlier state in the History pane. To preserve that history for reopening, choose File > Save RX Document. Keep an untouched original audio file too: plain Save can overwrite an opened WAV or AIFF.

When should I stop cleaning audio?

Stop when the defect no longer distracts and the recording still sounds natural. If another pass gives a small improvement but introduces obvious artifacts, return to the earlier, lighter version.