Short answer
Audacity can’t add detail that was never recorded, so the biggest improvement comes from a good recording. For a recording you already have, a standard speech chain in Audacity usually helps: a high-pass filter, light noise reduction, gentle EQ, compression, and then normalizing or loudness normalization. Export a lossless master.
What Audacity can and can’t fix
Audacity can:
- reduce steady background hiss or hum,
- cut low rumble and mic bumps,
- even out volume between loud and quiet parts,
- repair small defects, such as clicks.
Audacity can’t:
- restore detail that wasn’t captured. Upsampling, or converting an MP3 to WAV, adds no information.
- fully repair clipping. Clip Fix helps only with mild clipping.
- undo the damage from heavy processing. Heavy noise reduction produces watery, artifact-filled sound.
Start with the recording
- Record at 44.1 kHz or 48 kHz, in 24-bit. Audacity’s default project format is 32-bit float, which is fine for editing.
- Set the input level so peaks sit around -12 to -6 dBFS. That leaves headroom and avoids clipping.
- In Audacity 3.2 and later, the project sample rate is in the Audio Settings dialog (Audio Setup > Audio Settings).
Step 1: High-pass filter
Choose Effect > EQ and Filters > High-Pass Filter. For speech, a cutoff around 80 to 100 Hz removes rumble and mic thumps without thinning the voice much. The best value depends on the voice and the microphone, so preview a few settings.
Step 2: Noise reduction
- Select a few seconds of audio that contains only the room noise.
- Choose Effect > Noise Removal and Repair > Noise Reduction, then click Get Noise Profile.
- Select all the audio with
Ctrl+A. - Open Noise Reduction again. Start with a moderate amount, about 6 to 12 dB, and preview before you apply it.
Do this before compression. Compression raises the noise floor, so noise reduction works better when it comes first.
Step 3: EQ
Choose Effect > EQ and Filters > Filter Curve EQ, or use the Graphic EQ. Make small changes, a few dB at most. A gentle lift around 2 to 3 kHz can add clarity, and a small cut in the low-mids can reduce a boxy sound. Listen at normal volume after each change.
Step 4: Compression
Choose Effect > Volume and Compression > Compressor. A ratio of 2:1 and a threshold around -12 dB is a gentle starting point. Adjust until the level evens out without sounding pumped.
Step 5: Normalize or loudness normalization
- Normalize (Effect > Volume and Compression > Normalize) sets the peak level. It doesn’t reduce noise, and it raises any noise along with the audio.
- Loudness Normalization sets an average loudness, measured in LUFS. It’s the better tool when a platform specifies a loudness target. Many podcast guidelines aim around -16 LUFS for stereo, and streaming services often use around -14 LUFS, as described in the YouTube audio requirements. Targets vary, so check where you’ll publish.
Audiobook and broadcast distributors often set their own peak and loudness limits. For ACX’s, see audiobook recording software. Check them before you export.
Step 6: Export
- Save a lossless master as WAV or FLAC.
- For MP3, use 320 kbps constant bit rate, or a high-quality variable bit rate setting such as V0.
- Avoid repeated re-encoding. Export from the lossless master each time.
For the export dialog steps, see audio editing basics.
Common mistakes
- Expecting normalizing to make audio cleaner. It only makes it louder.
- Compressing before noise reduction, which raises the noise floor.
- Stacking many effects. Over-processing makes speech sound smeared.
- Upsampling or raising the bit depth, which adds no detail.
- Relying on Clip Fix for heavily clipped audio.
For clicks and other repairs, see removing clicks and pops. For silence at the ends of a recording, see trimming silence.