Breath remover

What the breath remover changes

Lower breaths toward the room tone

A breath is part of a performance, but a loud intake can distract from the words around it. The breath remover finds candidate events and lowers the selected ones toward the surrounding room tone. It does not cut time out of the recording or turn every intake into digital silence. That distinction helps a pause retain the texture of the original room rather than sounding like a hole in an edit. You can also leave a chosen amount above the room tone, so the target can be a softer natural breath instead of an entirely breathless voice. Listen through the whole pause before deciding.

Review the marks that matter

Automatic detection is a starting point, not an editorial verdict about how a person should breathe. The breath remover marks the events it found and lets you hear them individually. Untick an event to keep it when the breath carries emotion, signals the next phrase or turns out to contain useful speech. The detector is guarded against changing louder word sounds, but quiet consonants and unusual performances still need review. A convincing result preserves the rhythm of the speaker. Do not remove an intake simply because it has a mark; decide whether that particular sound is distracting in the final narration, conversation or spoken performance.

Keep the audio processing local

The detector model and the audio changes run in your browser. The breath remover downloads its processing code and model, then reads and saves the recording on your device rather than uploading it for a speech cloud run. Local processing can still take time to initialize, especially the first time the detector is needed. Keep the tab available until the result is ready. The detector was trained on English speech, so its scope should not be confused with a guarantee for every language or sung performance. The page provides the marks precisely so a person can review the places where automated choices matter.

When individual breathing decisions matter

• 

  Long-form narration

  In an audiobook or extended narration, breathing supports phrasing and helps a voice feel human. The breath remover can reduce a distracting intake without deleting the pause that gives the sentence its pace. Review the opening of a paragraph, a dramatic pause and a hurried phrase rather than accepting every marked event in a long file. A breath that sounds obvious in isolation may sit naturally between words at the final level. Use this workflow to address excessive breath prominence, not to erase all signs of performance. Keep a source copy and compare the processed passage over several sentences so the overall rhythm remains comfortable.

• 

  Podcast voice tracks

  A close microphone makes small inhalations prominent, especially once the voice is brought to podcast loudness. Use the breath remover when you want event-by-event control before a broader level workflow. Listen to each speaker separately if their microphone distances differ. One person may need almost no reduction while another has a few unusually loud intakes. The result should preserve a laugh beginning with an inhale and any breath that forms part of the reaction. For a complete spoken chain that also handles noise, clicks and level, compare the podcast audio enhancer, but retain the dedicated workflow when individual breaths need an explicit editorial decision.

• 

  Careful dialogue editing

  A dialogue editor may need the same pause length while lowering an audible inhale between lines. The breath remover keeps the timing and offers a softer event instead of a time cut. Compare the result under the picture when lip movement or body language makes the breath visible. A very dry pause can feel unnatural if the scene shows a person catching their breath. A detector does not decide the meaning of the performance. Keep or soften marks according to the scene, then check the transition across adjacent takes. Preserve the original so an important intake can be restored without trying to recreate it from a processed copy.

Detect, listen and choose

1. 1

   Load speech and initialize detection

   Choose the original spoken file and let the breath remover read it and initialize the detector. The first model load may take longer than later use. If the recording has a strong continuous noise floor, that noise can make breaths difficult to distinguish; read the page's explanation when the detector cannot confidently separate them. This is not a general background cleaner. Consider addressing the dominant noise from the source before evaluating breath control, and keep each stage as a separate copy. Start your review with a few clear intakes and a few quiet consonants so you learn how the detector behaves on this particular speaker.

2. 2

   Set the breath level and keep exceptions

   Listen to the marked events, then choose how close the selected breaths should sit to their surrounding room tone. The breath remover can leave them at that tone or somewhat above it rather than applying one universal silence cut. Untick any event that should stay original, including a mistaken mark or a deliberate expressive intake. Recheck the sentence after changing the amount. A result that sounds unobtrusive in isolation can still alter pacing emotionally, even when the timeline length is unchanged. Listen across a few sentences and retain enough breathing to keep the performance physically and conversationally believable.

3. 3

   Export after reviewing the performance

   Save the processed copy only after reviewing the marks that matter to your project. The breath remover is a targeted operation; it is not an automatic noise, echo or mouth-click cleanup chain. Listen again after final level processing, because quiet breathing can become more noticeable when the surrounding voice is raised. Keep the untouched source and a clear filename for the edited version. If a later review needs one expressive intake back, return to the source and adjust that mark. The breath remover makes event review practical, while the person editing remains responsible for deciding which parts of the recorded performance belong in the final piece.

Common questions