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

# Remove Wind Noise in iZotope RX 12: De-wind Guide
- URL: https://izotoperx.net/izotope-rx-remove-wind-noise/
- Published: 2026-09-04T07:54:21.000Z
- Updated: 2026-09-07T17:58:19.000Z
- Description: De-wind targets intermittent microphone rumble. Check overload first, tune the four controls to the recording, and preserve speech and outdoor ambience.
- Author: Brandon Hayes
- Tags: Repair Modules, RX 12, Dialogue Cleanup

**Use iZotope RX 12 De-wind when light or moderate gusts hit the microphone and create intermittent low-frequency rumble.** Select a representative burst with clean dialogue around it, use the visible wind as a starting point for Crossover Frequency and confirm the boundary by listening, then increase Reduction until the gust stops masking the voice. Add Fundamental Recovery only when the lower voice has thinned, and use Artifact Smoothing to trade watery musical noise against muffling.

De-wind is not the right first tool for constant background wind, broad hiss, or a blast that clipped the recording. RX's current manual says the module tracks changes in the noise floor and preserves steady background sound; heavy wind that blows out the microphone is outside its design. Diagnose the recording before processing: an unsuitable treatment can thin the voice without fixing the disturbance.

## Choose the right tool for the wind pattern

| What you hear and see                               | First tool                           | Why                                                |
| --------------------------------------------------- | ------------------------------------ | -------------------------------------------------- |
| Intermittent low rumble swelling with gusts         | De-wind                              | Matches the module's designed foreground pattern   |
| Constant broadband wind or steady outdoor roar      | Spectral De-noise                    | De-wind may preserve it as the noise floor         |
| One isolated leftover with usable surrounding audio | Spectral Repair                      | Targets only that time-frequency region            |
| Confirmed clipped peaks during a wind blast         | Test De-clip, then reassess          | Crackle alone does not establish clipping          |
| Dialogue buried under changing outdoor sound        | Dialogue Isolate                     | Separates voice from complex noise and reflections |
| Clothing rub or lav cable movement                  | De-rustle (Advanced) or local repair | Rustle is not wind rumble                          |

For recordings with several overlapping problems, follow the broader [background-noise cleanup guide](https://izotoperx.net/remove-background-noise-izotope-rx/) after identifying which sound is actually microphone wind.

## What RX 12 De-wind is designed to remove

The [current RX 12 De-wind manual](https://docs.izotope.com/rx12/en/de-wind.html?ref=izotoperx.net) defines the target precisely: intermittent low-frequency rumble caused by light to moderate bursts of wind contacting the microphone diaphragm. The module is marked ADV and Module, meaning RX Advanced and the standalone processing route rather than an ordinary real-time plug-in.

RX follows how the recording's noise floor changes over time and looks for the characteristic rise caused by a gust hitting the diaphragm. That is why De-wind can leave a steady outdoor bed alone while reducing foreground thumps and swells. It is also why selecting a long stretch of constant wind and expecting complete silence usually disappoints.

The [RX 12 edition and module tables](https://docs.izotope.com/rx12/en/rx-overview.html?ref=izotoperx.net) list De-wind only in Advanced, with no separate De-wind DAW plug-in. Standard includes the editor, EQ, Spectral De-noise, Dialogue Isolate and Spectral Repair, so it offers other ways to reduce wind. Elements is six DAW plug-ins with no RX Audio Editor; it does not include those four editor tools. An Elements user can try the host’s EQ and the included Voice De-noise, with De-clip reserved for actual overload. These are alternatives to test, not equivalents to De-wind.

## Intermittent gust, constant wind or clipping?

**Intermittent gusts** rise and fall against a more stable bed. In the spectrogram they often form dense low-frequency clouds or pulses. This is the De-wind case.

**Constant wind** stays present across the recording. It may sound like a broad roar, hiss or sustained low wash. RX's manual says De-wind can treat that as desirable noise floor and preserve it. A learned profile or Adaptive Mode in Spectral De-noise is usually the more relevant test.

**Clipped wind** has overloaded part of the recording chain. Listen for distortion and inspect flattened peaks; crackle alone can also come from handling or clothing. Clipping may sit below 0 dBFS after a level change, so use the waveform and De-clip histogram rather than only a red meter. A careful [De-clip test](https://izotoperx.net/fix-clipped-audio-rx-de-clip/) may soften clipping artifacts, but speech overwhelmed by the blast may remain unusable.

## How to see wind rumble in the spectrogram

Wind hitting a microphone usually concentrates energy toward the bottom of the spectrogram, but the exact boundary depends on the source, microphone, protection and gust. Look for low-frequency brightness that swells with the audible wind rather than setting Crossover from a memorized number.

Compare three moments: clean dialogue, a mild gust and the worst usable gust. If the bright region climbs into the voice's lower harmonics, aggressive reduction will have a cost. The [spectrogram guide](https://izotoperx.net/how-to-read-izotope-rx-spectrogram/) explains time, frequency and intensity so you can separate a wind cloud from a stable hum line or short handling thump.

Use headphones to identify the masking problem, but verify on speakers. Very low rumble may look dramatic and consume headroom without being obvious on small monitors; a higher broadband gust may remain audible after the low end is clean.

## Step-by-step De-wind workflow

1. Keep the original file untouched and open a separately named working copy in RX Audio Editor.
2. Check the waveform, playback level and De-clip histogram for actual overload; repair confirmed clipping on a test section before denoising.
3. Select a representative gust with intact dialogue before and after it, and identify the wanted channels.
4. Open De-wind from the module list in RX 12 Advanced and start with a restrained Reduction amount.
5. Use the upper edge of the wind cloud as a starting point for Crossover Frequency, then adjust it while listening to the voice.
6. Increase Reduction only until the gust is less distracting, keeping speech body and consonants intact.
7. Test Fundamental Recovery against nearby clean speech; keep only recovery that sounds natural.
8. Adjust Artifact Smoothing upward for watery residue or downward for muffled speech, and compare candidates with the untreated passage.
9. Render the chosen settings to the intended selection and channels, check the boundaries, then treat other affected passages and isolated leftovers individually.
10. Export a separately named audio file, reopen it and check the complete delivery in the destination session as well as by itself.

In the standalone module, click **Preview** to hear the selected passage and **Bypass** to alternate with the untreated audio. **Compare** stores a rendered candidate in the Compare Settings window; select it there and use Preview to audition it. The [RX module-control reference](https://docs.izotope.com/rx12/en/common-module-controls.html?ref=izotoperx.net) explains these separate actions. Neither Preview nor adding a Compare candidate commits the repair; **Render** does. Check the selection again before rendering, then use the [selection and History workflow](https://izotoperx.net/izotope-rx-selections-preview-compare-history/) to audition the changed boundaries or undo an unsatisfactory edit.

## Reduction: stop before the voice collapses

Reduction controls the balance between deeper wind suppression and preservation of the original signal. It is not a quality knob. Turning it farther can remove more rumble, but it also raises the chance of thinning speech or creating synthetic residue.

Begin low, then increase while the full sentence plays. Stop when the listener can follow the words and the gust no longer pulls attention. If one extreme burst remains after ordinary gusts sound good, leave the global setting alone and repair that event locally.

Judge at matched loudness. A processed clip can seem cleaner simply because low-frequency energy—and therefore overall level—fell. Bypass the module and compare the wanted voice, not just the noise.

## Crossover Frequency: follow the upper edge of the gust

Crossover Frequency sets the upper frequency limit for De-wind processing. Too low leaves wind energy above the boundary. Too high gives the algorithm access to more of the voice, ambience and other wanted sound.

Place the crossover around the top of the recurring wind cloud, preview, then move it in small steps. If consonants remain clear but the lower voice loses body, the issue may be Reduction or Fundamental Recovery rather than the boundary alone.

There is no universal crossover value for “male voice,” “female voice” or “shotgun mic.” A memorized preset ignores the actual wind spectrum. Read the selected recording.

## Fundamental Recovery: restore body only when needed

According to the current De-wind manual, Fundamental Recovery re-synthesizes lower voice harmonics lost or obscured by wind. Treat it as a reconstruction control, not ordinary bass EQ. If reducing wind leaves the speaker sounding small or papery, compare a little recovery with both the untreated passage and nearby clean speech.

Increase it cautiously and listen for a plausible match to the same speaker on the same microphone. Stop or back it off if the low end becomes artificial. A missing clean reference makes this judgment less certain; it does not justify turning recovery up until the signal looks fuller.

This control is designed around speech fundamentals. Do not assume it will rebuild missing bass, engine texture or musical low end with the accuracy of an original recording.

## Artifact Smoothing: watery versus muffled

De-wind uses FFT-based processing, and strong removal can produce “musical noise”: moving, underwater-like tones in the residue. The manual says to increase Artifact Smoothing when the output sounds watery, but decrease it when smoothing makes speech muffled.

Tune it after Reduction and Crossover are close. If you smooth first, you may hide the symptom while leaving an overly aggressive reduction. Listen through the whole gust and its release; artifacts often appear as the noise floor returns.

The manual also warns that removing more source audio makes undesirable FFT artifacts more likely. A second restrained pass can help some constant or residual cases, but it can also compound damage. Compare it against one lighter pass plus a local edit.

## De-wind vs Spectral De-noise

Choose De-wind for intermittent foreground gusts that strike the diaphragm. Choose Spectral De-noise for a sustained or background wind bed. This distinction comes directly from the current manual rather than from marketing shorthand.

For a steady bed, select noise without speech or other wanted sound, click **Learn** in Spectral De-noise, then select the passage you intend to process. The learned profile stays fixed. If no usable noise-only section exists or the bed changes, test **Adaptive Mode**, which updates the profile from incoming audio. The [current Spectral De-noise guide](https://docs.izotope.com/rx12/en/spectral-de-noise.html?ref=izotoperx.net) covers both routes. A gust with words in it is not a safe noise-only example: learning those words can make the process remove them.

If both patterns are present and clipping has been addressed, try reducing the dominant gusts, then reassess whether the steady remainder needs another tool. Keep the second stage only when it improves the complete sentence. The manual also allows trying De-wind more than once, but that is an experiment to compare against a lighter single pass, not a required chain.

## Use Spectral Repair for isolated leftovers

A single gust that survives an otherwise useful De-wind pass should not force harsher settings across the whole file. Select only the unwanted time-frequency area in Spectral Repair. Leave useful neighboring audio outside that selection: the module reads its reference from the **Surrounding Region**, whose extent you adjust separately. Including clean surroundings inside the damaged selection makes them part of the repair target.

Start by comparing Attenuate when wanted speech remains audible under the noise. Replace substitutes the selected content with an estimate from surrounding audio, so a missing word is not automatically recoverable. The [Spectral Repair mode reference](https://docs.izotope.com/rx12/en/spectral-repair.html?ref=izotoperx.net) limits Replace and horizontal or 2D Attenuate selections to 10 seconds; keep these local edits short and verify the active mode. The [Spectral Repair tutorial](https://izotoperx.net/izotope-rx-spectral-repair/) explains selection, weighting and artifact checks in more detail.

Older official material, including iZotope's [dialogue-repair case study](https://www.izotope.com/community/blog/6-dialogue-repair-problems?page=5&ref=izotoperx.net), shows why wind once required several tools. Use those examples as workflow history, not proof that RX 7 screenshots match RX 12.

## What to do with a clipped wind blast

Once clipping is confirmed, test De-clip on a duplicate before De-wind. The [current De-clip controls](https://docs.izotope.com/rx12/en/de-clip.html?ref=izotoperx.net) place the Threshold just below the clipping level shown by the selected audio, rather than at an arbitrary 0 dBFS. Reconstructed peaks can rise, so use Makeup Gain or the Post-limiter as appropriate and check the result for new clipping. De-clip can reduce overload distortion; it cannot restore a speech waveform that the microphone never captured usefully.

After De-clip, reassess the spectrogram. If intermittent rumble remains, apply De-wind conservatively. If the word is still missing, an alternate microphone, ADR or an audio edit covered by a cutaway may be more honest than aggressive synthesis.

A 32-bit float file above 0 dBFS can overload playback while retaining usable samples. As [Sound Devices explains](https://support.sounddevices.com/hc/en-us/articles/41709146697243-Understanding-32-bit-Float?ref=izotoperx.net), attenuation can bring intact over-zero samples into range. Test that level correction before treating them as damaged peaks. This does not undo distortion already produced in the microphone or analog chain. For genuinely clipped material, boosting or normalizing is not a substitute for repair; bring the final level into range before delivery.

## Dialogue Isolate when wind overlaps the voice

Dialogue Isolate can be more useful when the problem extends beyond low-frequency diaphragm rumble and the voice sits inside complex, changing outdoor sound. It separates dialogue from noise and reflections rather than tracking only the wind signature.

Use it gently after diagnosing clipping and gusts. In Standard and Advanced, Dialogue Isolate separates Voice, Noise and Reverb; increasing Sensitivity classifies more material as noise or reverb and can damage speech. The [current Dialogue Isolate manual](https://docs.izotope.com/rx12/en/dialogue-isolate.html?ref=izotoperx.net) reserves the Quality selector and four-band processing for Advanced in both the module and plug-in. Best / Offline is intended for the editor and offline DAW renders. The [Dialogue Isolate guide](https://izotoperx.net/izotope-rx-dialogue-isolate/) covers the wider cleanup workflow.

Do not stack aggressive De-wind, Spectral De-noise and Dialogue Isolate by default. Each stage sees artifacts from the previous one. Keep a stage only when its processed version improves on bypass.

## Keep an outdoor scene sounding outdoors

For a scene meant to sound outdoors, studio-dry dialogue can expose the repair. Remove distracting microphone turbulence while retaining a coherent weather bed. When a cut reveals a gap, use suitable recorded ambience to restore continuity; the amount is an editorial choice for the scene.

Check edit boundaries, perspective and the other production microphones. A lav, boom and camera track may contain different wind signatures. Repairing one until it is unnaturally clean can make the transition more obvious.

The older [official De-wind introduction](https://www.izotope.com/en/learn/remove-wind-noise-from-audio-with-de-wind-in-rx-7.html?page=4&ref=izotoperx.net) correctly frames the task as reducing intermittent rumble while retaining dialogue intelligibility; only its RX 7 interface context is historical.

## Prevent the next wind problem on location

Reduce airflow at the microphone before relying on software. [RØDE’s wind-protection guidance](https://rode.com/en-au/about/news-info/audio-for-film-101-ways-to-reduce-wind-noise-when-recording-outside?ref=izotoperx.net) distinguishes foam for light wind from furry windshields and enclosing blimps for stronger conditions. Use protection designed for the microphone, fit it correctly, and test the actual camera or boom movement. Shelter or a change in position may help, provided the voice still sounds right and clothing does not rub the mic.

Monitor on closed headphones and record a rehearsal in the actual position. Test both quiet dialogue and likely gusts. Keep appropriate input headroom for the recorder and microphone: a lower digital level cannot protect a capsule that is already overloaded. Check alternate production microphones separately instead of assuming a clean meter means every channel is usable.

Where the production allows, record a second protected microphone and enough uninterrupted location ambience to cover the expected edits. The alternate track may supply a damaged word, and the ambience can smooth repaired cuts. Mark unusable takes while a replacement recording is still possible.

Music and nature recordings deserve an extra low-end check. In a [June 2023 Taperssection discussion](https://taperssection.com/index.php?topic=202629.0&ref=izotoperx.net), a recordist reported losing wanted bass while reducing wind and limiting the treatment to dialogue between songs. That historical report is not an RX 12 benchmark; it illustrates why speech-focused recovery and broad low cuts should not be assumed to preserve a concert’s bass.

## De-wind troubleshooting

| Symptom                      | Likely cause                                  | First fix                                         |
| ---------------------------- | --------------------------------------------- | ------------------------------------------------- |
| Steady wind remains          | De-wind preserves the constant noise floor    | Test Spectral De-noise                            |
| Dialogue loses body          | Too much Reduction or insufficient recovery   | Reduce less, then test Fundamental Recovery       |
| Output sounds underwater     | FFT musical noise                             | Raise Artifact Smoothing slightly or reduce less  |
| Speech becomes muffled       | Too much smoothing or crossover reach         | Lower smoothing and re-check Crossover            |
| One blast still crackles     | Possible clipping, handling or remaining wind | Confirm the cause; test De-clip only for clipping |
| De-wind is missing           | Wrong edition or expecting a plug-in          | Confirm RX 12 Advanced standalone module          |
| Background pumps after gusts | Over-processing or a changing noise bed       | Back off and preserve the bed                     |

## Final quality-control checklist

- The problem is intermittent diaphragm wind, not constant noise alone.
- Clipping was checked before denoising.
- Crossover follows the observed gust rather than a copied preset.
- Reduction stops when the voice is clear, not when the spectrogram is empty.
- Fundamental Recovery was compared with available clean speech and reduced if it sounded artificial.
- No new watery artifacts, muffled consonants or unacceptable speech damage were introduced.
- Isolated leftovers were repaired locally.
- Outdoor ambience stays continuous and believable.
- The complete exported result was checked at matched voice level outside RX, including cuts and quiet sections.
- The original recording remains available.

For delivery, use **File → Export…** to write a separately named WAV or AIFF with the intended format, channel layout and duration. **Export Selection…** writes only the current selection. Reopen the complete output, check peaks and listen through it in sync with picture where relevant; a clean test excerpt is not final QC. Save an .rxdoc separately if you need the editable history. The [RX file-saving reference](https://docs.izotope.com/rx12/en/working-with-files.html?ref=izotoperx.net#saving-files) warns that ordinary Save on WAV or AIFF overwrites the source; the [import and export guide](https://izotoperx.net/izotope-rx-import-export-save/) explains delivery choices. For other defects, the [Repair Audio hub](https://izotoperx.net/repair-audio-rx/) helps identify the relevant repair method.

## Frequently asked questions

### What are the best RX 12 De-wind settings?

There is no universal preset. Set Crossover near the upper edge of the actual gust, raise Reduction only until masking stops, add Fundamental Recovery if speech thins, and balance Artifact Smoothing between watery and muffled.

### Is De-wind included in RX Standard?

No. The current RX 12 manual and feature comparison list De-wind as an Advanced-only module.

### Can De-wind remove constant background wind?

Not reliably. De-wind tracks noise-floor changes and may preserve constant wind as part of the background. Spectral De-noise is the documented alternative.

### What does Crossover Frequency do?

It sets the upper frequency limit for De-wind processing. Set it high enough to cover the gust but not so high that the module reaches unnecessary voice and ambience.

### What does Fundamental Recovery do?

It re-synthesizes lower voice harmonics lost or obscured by wind. Compare cautiously with nearby clean speech; it is an estimate, not guaranteed recovery of missing original audio.

### Why does De-wind sound watery?

Strong FFT-based removal can create musical noise. Increase Artifact Smoothing slightly, reduce the processing depth, or use a smaller local repair for the worst gust.

### Can RX repair wind that clipped the microphone?

De-clip may reduce overload distortion, but De-wind is not designed for heavy blasts that blow out the signal. Speech completely masked or destroyed may not be recoverable.

### Should I use De-wind or Dialogue Isolate?

Use De-wind for intermittent low-frequency gusts hitting the diaphragm. Use Dialogue Isolate when speech is buried in broader, changing noise and reflections.