How to Clean Audio from a Microsoft Teams Meeting Recording

A Microsoft Teams meeting recording can be a goldmine: a podcast episode, a training video, an interview, or a webinar replay. It can also sound terrible—laptop fans, keyboard clatter, room echo, notifications, and a participant on a bad headset. The good news is you can clean it up after the fact. You do not need the original meeting or a re-record, just a sensible order of operations and a few tools. This guide walks through cleaning audio from a Microsoft Teams meeting recording without destroying the voices.

First, Understand What You Are Working With

Teams stores most meeting recordings in OneDrive or SharePoint as an MP4. The audio is usually mixed into one stereo track, not separated per participant, so one noisy person affects the whole file. Download the original recording—do not screen-record playback. A transcript or captions file can help with editing words, but it does not remove noise. For cleanup, you want the original MP4 and a clean WAV extraction.

Step 1: Extract the Audio from the Teams MP4

Work on a WAV file, not the MP4. WAV is uncompressed, so edits are less destructive. Use ffmpeg if you are comfortable with the command line:

ffmpeg -i TeamsRecording.mp4 -vn -acodec pcm_s16le -ar 48000 -ac 1 teams_audio.wav

That drops the video and creates a 48 kHz, 16-bit mono WAV. Mono is usually better for voice noise removal because it simplifies processing and most AI models expect it. If you need stereo for music or spatial cues, use -ac 2. Tradeoff: stereo can make noise reduction less predictable. No command line? Open the MP4 in Audacity and export as WAV, or use Shutter Encoder. Keep the original MP4 as a backup.

Step 2: Identify the Noise Before You Remove It

Put on headphones and listen once. Mark timestamps for major problems. Not all noise is the same.

• Constant noise: fan hum, air conditioning, computer whine, room tone. This responds well to noise reduction.
• Transient noise: keyboard clicks, mouse clicks, coughs, notification chimes. These need manual repair or de-click tools.
• Echo or reverb: hollow, slap-back sound from a bad room or speakerphone. Noise reduction will not fix this. You need de-reverb or a re-record.
• Clipping or distortion: crackly peaks. You can reduce it slightly, but clipped audio cannot be fully restored.

If only one participant is noisy, process that section separately. A global noise reduction pass can make the clean speaker sound hollow. Tradeoff: section editing takes longer, but it sounds more natural.

Step 3: Start with Gentle Cleanup

Before aggressive noise removal, fix the low end and any electrical hum. In Audacity, select the whole file and use Effect > High-Pass Filter at 80–100 Hz with a 12 dB per octave rolloff. This removes rumble, desk bumps, and low fan noise without touching most speech.

If you hear steady hum, add a notch filter. In North America, hum is usually 60 Hz; in Europe and many other regions, it is 50 Hz. Use Effect > Notch Filter with a high Q (10–20). If the hum is strong, also notch the first harmonic (120 Hz or 100 Hz). Tradeoff: too much notching can thin the voice. Apply only if you actually hear hum.

Step 4: Reduce Noise with a Noise Profile or AI Tool

Manual route: Audacity Noise Reduction

1. Find 0.5–1 second of room tone—no speech, no keyboard, just background noise.
2. Select that section and choose Effect > Noise Reduction > Get Noise Profile.
3. Select the whole file, open Noise Reduction, and start with Reduction at 12 dB, Sensitivity at 6, and Frequency smoothing at 3.
4. Preview. If the voice sounds watery or metallic, lower Reduction to 8–10 dB. If noise remains, raise Sensitivity slightly, not Reduction.

Tradeoff: too much reduction creates artifacts and dulls consonants. A little remaining fan noise is often better than a robotic voice.

Automatic route: AI noise removal

For hour-long webinars or interviews, an AI tool is faster. Adobe Podcast Enhance is one option but has upload limits. A free online option like ClearVoiceHQ (clearvoicehq.com) is built for noise removal in remote meetings, webinars, and interview recordings. Upload the extracted WAV or original MP4, run noise removal, and download. Tradeoff: automatic tools give less control. Test on a 30-second clip first. If it over-processes, try a gentler setting or blend the cleaned audio with the original at 20–30%.

Step 5: Repair Clicks, Coughs, and Other One-Off Sounds

AI noise reduction handles constant noise, not sudden clicks. In Audacity, zoom in until you see the click waveform. Select just the click—usually 1–5 milliseconds—and use Effect > Repair. For coughs or chair squeaks over silence, replace the sound with nearby room tone or mute it. Use short crossfades so you do not hear a gap. Tradeoff: repeated repairs can soften the audio, so use them sparingly.

Step 6: Add EQ, Compression, and Loudness

After noise removal, the voice may sound dull. Add gentle EQ. If the voice is muddy, cut 200–400 Hz by 2–3 dB. If it is harsh, cut 2–4 kHz by 1–2 dB. If it lacks presence, boost 3–5 kHz by 1–2 dB. Keep moves small.

Then compress. For speech, start with ratio 2:1 to 3:1, threshold around -18 dB, attack 10 ms, release 100 ms, and makeup gain to bring the level back up. The goal is consistent volume, not a crushed podcast sound.

Finally, normalize loudness. For meetings and podcasts, target -16 LUFS integrated with a true peak of -1.5 dBTP. For YouTube, -14 LUFS is common. Use a free loudness meter. Tradeoff: over-compression brings up residual noise, so clean first, compress second.

Step 7: Put the Clean Audio Back into the Video

If you need the video, replace the original audio with your cleaned WAV. With ffmpeg:

ffmpeg -i TeamsRecording.mp4 -i cleaned_audio.wav -map 0:v -map 1:a -c:v copy -c:a aac -b:a 192k final_output.mp4

This copies the video without re-encoding and adds the cleaned audio as AAC. If you trimmed or shortened the audio, check sync before exporting. In a video editor, drag the cleaned WAV onto the timeline and align it with the first frame. Export at 192 kbps or higher.

Teams-Specific Tips for Future Recordings

• Before recording, set Teams noise suppression to High: Settings > Devices > Noise suppression.
• Use a headset or USB microphone instead of a laptop mic.
• Ask everyone to mute when not speaking and silence keyboard notifications.
• Record locally with separate tracks if possible. OBS can capture each participant on its own track, so you can clean each voice individually.
• For important interviews, have each side record locally as a backup.

Which Tool Should You Use?

• Short clip, free, full control: Audacity.
• Long meeting, fast turnaround: ClearVoice or Adobe Podcast Enhance.
• Severe echo or music: iZotope RX or Adobe Audition.
• Batch processing: Auphonic or Descript.

No tool is perfect. Start with high-pass, then noise reduction, then de-reverb. Compare before and after on headphones, a phone speaker, and a laptop speaker. If it sounds worse, undo and try a gentler setting.

Final Checklist

1. Download the original Teams MP4 from OneDrive or SharePoint.
2. Extract a 48 kHz mono WAV.
3. High-pass at 80–100 Hz.
4. Notch 50/60 Hz only if you hear hum.
5. Apply noise reduction with a room-tone profile or AI tool.
6. De-reverb if the room sounds hollow.
7. Repair clicks and coughs manually.
8. EQ and compress gently.
9. Normalize to -16 LUFS for speech, -14 LUFS for YouTube.
10. Recombine with video and check sync.
11. Listen on at least two playback systems.

You cannot fully undo a bad Teams recording, but you can make it clean and listenable. The order matters: repair first, reduce noise second, enhance last. For the fastest route on a long meeting or webinar, try an automatic tool like ClearVoice, then do a quick manual polish. Keep a backup of the original file, and you can always start over.