You have finished recording a tutorial and can hear a fan underneath your voice. Before applying a filter to the whole video, check whether the microphone was saved on its own track. A separate voice track gives you a way to try noise reduction without deliberately processing the system sound.
If everything was mixed together, make a copy and test a short section. Speech denoising may also change music, alerts or game audio on that track; it cannot reliably recover separate sources from the mix.
First, check what is on each track
Open a copy of the recording in an editor that can display its audio tracks. A media player may play only one track, so hearing just the voice or system audio in a player does not by itself tell you what the file contains. Look for separate tracks in the editor, then listen to each one in isolation.
The OBS multitrack guide explains why track selection and container support matter. It describes settings for future recordings; changing those settings cannot create separate tracks in an already mixed file.
Export a short microphone clip
- Duplicate or save a copy of your editing project. Keep the original recording untouched.
- Select a representative passage on the microphone track, including speech and a brief pause. Note where the selected passage starts on the timeline.
- Export that selection as a separate supported audio file. Use a compatible WAV, MP3 or AAC-in-M4A file; a WAV must be 16-bit PCM, mono or stereo, at 44.1 or 48 kHz.
- Check both input limits: 25 MiB and five minutes. If the excerpt is too large, select a shorter passage in the editor.
The exact controls depend on your editor. For example, Audacity documents exporting selected audio. Check the chosen track and export range so you do not accidentally export a mix of the entire project.
Compare the original and cleaned voice
Open the audio noise remover, choose the exported clip and create a preview. The preview contains up to the first 30 seconds. Listen to the same sentence in Original and Processed at a consistent volume. Pay attention to quiet word endings, breaths and the point where speech starts; a quieter pause is not useful if the explanation becomes harder to understand.
If the clip is longer than the preview and within both limits, select Process full audio before downloading the full WAV. For a shorter clip the first pass covers the entire clip. The tool has no adjustable strength control. If it damages the voice, retain the original or use an editor with the controls you need.
Put the WAV back without shifting the video
Import the downloaded WAV into your editor on a new audio track. Align its start with the start of the section you exported. Temporarily mute the corresponding original microphone section and play across the beginning and end of the replacement. Check lip sync when a face is visible, and listen for gaps, abrupt level changes or two copies of the voice playing together.
Leave the system-audio track in place. Save a new project version and export a test from your editor. Listen to the complete mix as well as the isolated voice. The browser tool exports audio; you still need to replace the track in your editor.
What if the voice and system sound are mixed together?
Try a duplicate only if you are willing to compare all the sounds in it. A speech-first process may reduce or alter the system audio. If preserving those sounds is essential, look for an earlier separate microphone recording, a project backup or a different edit. Do not promise that a denoiser will reliably separate the sources.
The separate video workflow on this site has its own compatibility limits and exports a preview of up to 30 seconds. It is not a way to export an entire long tutorial. If you need to preserve the full picture, use your video editor for the final export.
Screen recording questions
Can I remove keyboard clicks too?
Individual clicks are not the same as steady fan noise. They may remain, and this tool does not promise click repair.
Will the cleaned WAV stay in sync automatically?
No. Place it at the matching start point and listen across the joins in your editor.
Can I use this for a full hour-long recording?
Not as one audio file: both the five-minute and 25 MiB limits apply. A full editor is the practical place to finish a long project.