A voiceover with a steady fan underneath it may be worth cleaning before you finish a CapCut project. First check the options in your current CapCut version: its features, interface and access can vary by platform and change over time. If the native workflow already gives you the result you need, use that.
The alternative below is for a separate voiceover audio file. It is not a method for reliably separating speech, music and sound effects that have already been mixed together.
Find the original voiceover
If you have the microphone recording as its own file, keep it untouched and work on a copy. Find a representative passage with speech and a pause. If the voice is embedded only in a mixed video, listen to the whole mix before deciding to process it: a speech-focused denoiser may change music or effects as well as the noise.
Do not rename a video to an audio extension. If you need an audio-only copy, export it from an editor in a supported format. The relevant export controls depend on the editor and version; check that the selected range and track contain what you expect.
Make a short preview
- Prepare a WAV, MP3 or AAC-in-M4A voice clip no longer than five minutes and no larger than 25 MiB. Both limits apply. Supported WAV files are 16-bit PCM, mono or stereo, at 44.1 or 48 kHz; ALAC is unsupported.
- Open voice recording cleanup, choose the file and create a preview of up to the first 30 seconds.
- Listen to Original and Processed at a consistent volume. Check that the explanation remains clear and that breaths and consonants are not cut off.
- If the clip is longer than the preview and the result is acceptable, process the full supported audio and download the WAV.
This tool has no strength slider. If the result sounds watery or loses words, keep the original voiceover.
Place the WAV back in the project
Import the downloaded WAV onto a new audio track and align it with the start of your exported selection. Mute or replace the matching original section so you don't hear two voices. Leave separate music and system-audio tracks in place.
Play across the first and last edit points. Check the speaker's mouth if visible, then listen for doubled words, gaps or sudden level changes. Save a copy of the project before replacing anything permanently. Export a short test from your editor and inspect it before making the final video.
What if there is no separate voice track?
If the only source is a finished mix, the browser tool may also alter background music and sound effects. You can test a copy, but do not assume it will preserve every non-speech sound. Look for the original microphone recording or an earlier project with separate tracks. If neither exists and the mix sounds worse after processing, retain it or seek a specialized editing workflow.
Questions about this workflow
Does this replace CapCut?
No. It prepares a supported audio clip; the timeline and final video export remain in your editor.
Does CapCut charge for noise reduction everywhere?
Availability and pricing can differ by version, platform and account. Check the controls in the version you actually use.
Can this clean all of a long video?
Not as a single audio file beyond the five-minute and 25 MiB limits. The separate video mode has different restrictions and exports at most a 30-second preview.