Noise reduction is not binary (on or off). It operates on a spectrum from subtle cleanup to aggressive suppression. The right setting depends on your content type, the severity of the noise, and where the audio will be consumed.
Light reduction (10-30% strength)
Use for: Podcasts, interviews, narration where speech quality is the top priority.
What it does: Reduces the noise floor by 6-10 dB. The hum or hiss becomes noticeably quieter but does not disappear entirely. Speech remains completely natural with zero artifacts.
When to choose this: When the noise is mild (quiet fan, faint hiss) and the audio will be listened to with full attention (podcast, audiobook, voiceover). Listeners will tolerate a faint background noise if the speech sounds natural. They will not tolerate robotic or watery speech even if the background is silent.
Moderate reduction (40-60% strength)
Use for: Tutorial videos, screen recordings, vlogs where audio is important but not the sole focus.
What it does: Reduces the noise floor by 12-18 dB. Most steady-state noise becomes inaudible. Speech may have very subtle artifacts (slight thinning of sibilants, minor loss of breathiness) that only trained ears notice.
When to choose this: When the noise is moderate (loud fan, audible AC) and the audio accompanies video. Viewers are watching the screen, so they are less attentive to subtle audio artifacts. The reduction of distracting noise improves the overall viewing experience.
Aggressive reduction (70-100% strength)
Use for: Social media clips, short-form content, situations where the alternative is muting entirely.
What it does: Reduces the noise floor by 20-30+ dB. Background noise is essentially eliminated. Speech may sound noticeably processed: metallic timbre, loss of room ambience, audible "pumping" (volume fluctuations as the algorithm engages and disengages during pauses).
When to choose this: When the noise is severe (loud machinery, heavy traffic) and the audio would be unusable without aggressive processing. The artifacts of aggressive reduction are preferable to muting the video entirely, but the audio will not sound professional.
Content-specific recommendations
YouTube videos (educational, tutorial): Moderate. Viewers watch these with headphones and notice noise, but also notice aggressive artifacts. Find the middle ground.
TikTok and Instagram Reels: Moderate to aggressive. Playback is on phone speakers in noisy environments. Subtle noise is masked by ambient sound. Aggressive artifacts are also masked. Cleaner audio stands out on the platform.
Podcasts: Light. Podcast listeners are audio-focused and use headphones. They will notice any processing artifacts. Natural-sounding speech with mild background noise is strongly preferred over processed speech with no noise.
Corporate presentations: Moderate. Clean enough to be professional, natural enough to not distract from the content.
Testing your settings
Listen to the processed audio on multiple playback devices before publishing:
1. Headphones (reveals artifacts most clearly) 2. Phone speaker (reveals whether noise is still audible on low-quality playback) 3. Laptop speakers (the most common viewing scenario for many content types)
If the audio sounds good on all three, your settings are right. The MediaBrew audio denoise tool lets you adjust the strength and preview results before committing. Process locally via FFmpeg WebAssembly.