Audacity Noise Reduction Sliders Explained

Audacity's Noise Reduction effect uses three primary sliders—Noise Reduction (dB), Sensitivity, and Frequency Smoothing (bands)—to isolate and reduce unwanted background sound from a recording. By capturing a "Noise Profile" first, the software identifies unwanted frequencies, and these three controls determine how much volume to remove, how strictly the software identifies the noise, and how smoothly the frequencies are blended to prevent digital distortion.

Noise Reduction (dB)

The Noise Reduction (dB) slider controls the volume attenuation applied to the identified noise. Measured in decibels, this setting dictates how quiet the background sound becomes once the filter is applied.

• Higher values: Attenuate the noise more aggressively. If set too high (above 12 to 18 dB), it often introduces a hollow, robotic, or underwater sound to the remaining audio.
• Lower values: Apply subtle reduction, lowering the noise floor without compromising the natural tone of your main audio.
• Typical setting: Between 6 dB and 12 dB is generally sufficient for voice recordings.

Sensitivity

The Sensitivity slider controls the threshold used to distinguish between unwanted background noise and the sound you want to keep. It acts as a bias control for the noise detection algorithm.

• Higher values: Make the filter more aggressive, categorizing more sounds as noise. This can eliminate faint, fluctuating background sounds, but it risks cutting into the lower frequencies or tails of words in vocal tracks.
• Lower values: Make the filter conservative, ensuring that your primary audio remains completely untouched. However, this may leave audible background hiss or hum behind.
• Typical setting: A value between 6 and 8 provides a balanced detection threshold for most microphones and acoustic environments.

Frequency Smoothing (bands)

The Frequency Smoothing (bands) slider controls how the reduction is blended across neighboring frequency bands. When noise is subtracted, it can leave behind random, isolated frequency spikes known as "spectral artifacts" or "musical noise," which sound like faint digital chirping or metallic ringing.

• Higher values: Spread the noise reduction across adjacent frequency bands, smoothing out the transition and eliminating chirping artifacts. However, setting this too high can reduce the overall clarity and crispness of your audio.
• Lower values (or 0): Keep the reduction strictly targeted to the exact frequencies detected in the noise profile, preserving maximum vocal clarity. Setting it to 0 is common for voice tracks, provided musical noise does not become noticeable.
• Typical setting: Set to 0 to 3 for spoken-word content, and up to 6 for music or complex soundscapes where artifacts are more noticeable.