How to Remove Wind Noise from Video Without Losing Speech

To remove wind noise from a video, first listen to a gust that overlaps speech. Low rumble may respond to a high-pass filter; changing wind around an intelligible voice may be a candidate for voice isolation. If a gust has already distorted or obscured a word, a quieter background does not prove that the word has been recovered.

Keep the original video and compare a short, representative passage before treating the entire recording. Include a calm sentence, the strongest gust, a quiet word and the return to calmer sound. Testing only a pause can hide damage to the message you need to keep.

Identify the wind problem before choosing a filter

Starting checks for different outdoor-recording problems

What you hear	First comparison	What can go wrong
Low, continuous rumble under clear speech	Preview a conservative high-pass filter on a copy	The voice can lose body along with the rumble.
Changing rushes around otherwise understandable words	Test voice isolation from the original excerpt	Speech can acquire artifacts or lose quiet syllables.
A loud burst makes a word crackle or disappear	Check the original and any alternate microphone track	Separation cannot be assumed to reconstruct missing speech.
Only a pause is distracting	Lower that interval in the editor, preserving its duration	Deleting the time moves everything that follows out of sync.
The whole clip sounds strange on one playback device	Compare the original through another available playback path	A playback problem can be mistaken for recording damage.

Check for a separate microphone recording before repairing the camera mix. If one is available, audition it at the matching source time. It may contain a different balance of wind and speech. Two channels in a stereo file do not automatically mean separate microphones: listen to both before deciding that one side is a cleaner source.

Try a high-pass filter for low wind rumble

A high-pass filter progressively reduces lower frequencies. It does not recognize wind or separate the voice from other sounds. Audacity's current High-pass Filter reference places it under Effect → EQ and filters. The controls are cutoff frequency and roll-off; use Preview before applying.

The manual gives 60–100 Hz as a speech-rumble starting range. Treat that as a comparison range, not a setting guaranteed to suit every voice. Its documented 1,000 Hz default is not a reason to apply that cutoff blindly to dialogue. Raise or lower the cutoff while checking both the rumble and the speaker's body; use your version's manual if controls differ.

Compare each setting from the original at a similar listening level. If the voice becomes thin before the wind is acceptably reduced, return to the original and test another method.

A low-cut treatment is not a repair for every gust. Wind can cover more than the lowest frequencies, and a distorted burst can still sound unpleasant after its bass is reduced. A less boomy recording can remain unintelligible.

To test a low cut without installing an editor, the free local high-pass filter accepts short MP3/WAV files and exports a real WAV or MP3. Its slope is fixed at 12 dB/octave; it does not process a video directly or recognize wind. Keep the original and compare the voice as well as the rumble.

Test voice isolation when the words are still audible

The wind-noise removal workflow processes an uploaded recording. For dialogue, start with a description such as “The person speaking in the foreground.” Review Isolated sound for the requested voice and compare it with the original. Background audio is the remaining mix, not the automatic choice for keeping speech.

1. Choose a representative source. Include the gust and complete surrounding words. Save the source start time if you make an excerpt.
2. Upload the supported file. Saturalabs accepts MP4 or MOV video, or MP3/WAV audio, up to 50 MiB. Its free audio trimmer can prepare a short supported audio excerpt; it does not repair the wind itself.
3. Request one audible voice. Use a description that fits the recording. A generic voice target can retain several people; it is not a promise of one track per speaker.
4. Review the difficult moment. Follow the sentence without reading a transcript. Check whether its beginning, quiet middle and ending remain intact during the gust.
5. Keep only a useful candidate. Compare the actual output before a WAV download. Downloads require an account and sufficient credits; the results screen shows the cost.

If the outdoor ambience matters as well as the voice, a separate wind-targeted attempt from the original can be another candidate. Describe the audible unwanted wind and review Background audio for the remainder. Do not assume this preserves every tree, bird, footstep or audience sound. Choose by the scene you need, and keep the two requests and output names clearly labeled.

Why a noise profile or a gate can disappoint

Audacity's Noise Reduction reference describes profiling an isolated noise sample, then applying reduction to selected audio. It suits a steady background. The manual warns that wanted sound included in the profile can be removed too, and recommends listening to the Noise only output to check the material being taken away.

A changing gust is not automatically represented by a short calm pause. If the noise character changes through the take, a profile that sounds useful in one place may damage speech elsewhere. Do not label an interval “wind only” if it contains a quiet word or an important scene sound.

The local noise gate lowers quiet passages by level. A loud gust can open it, while wind mixed beneath louder speech can remain. The silence remover instead shortens gaps; it does not recognize wind, and its output should not replace an unchanged video's soundtrack. The cleanup-method guide compares these operations in more detail.

Return the chosen audio to the video without changing timing

Import the WAV into your video editor and align it with the source passage. For a full-source export starting at zero, use the corresponding source start. For an excerpt from source 42.5–68 seconds, place it at that excerpt's position in the project, not automatically at timeline zero. Follow the audio replacement guide when the original clip has already been trimmed or moved.

Mute the corresponding original audio while comparing the replacement. Leaving both versions audible can produce doubled speech and a misleading impression of leftover wind. Preserve the original as a separate reference, and check a visible speech cue near both ends after the final video export.

Review the transition into and out of any repaired excerpt. A sudden change from outdoor room tone to near silence can draw attention to the edit. Use your editor's available boundary controls deliberately; an outer fade on an entire file does not automatically repair an internal replacement boundary.

Decide when another take is the better repair

Keep a specific note such as “the word after the gust at 00:18 is still missing,” rather than only “less noisy.” If the sentence cannot be understood from the available recording, compare another take or re-record narration when that fits the project. Captions can accompany an edit, but they do not restore the original spoken audio.

For the next recording, follow your microphone's wind-protection guidance and make a short outdoor playback test. RØDE's VideoMic GO II guidance, for example, distinguishes the supplied foam for light wind from stronger-wind protection for that microphone. That is a compatibility example, not a recommendation that one accessory fits every camera or phone.

When practical, compare a more sheltered speaking position and a microphone closer to the intended voice, while maintaining an appropriate recording level. Record both a normal sentence and a quiet ending under the actual conditions. Listening to that short take before continuing is more informative than assuming a filter will solve every gust later.