Speech workflows
Clean up a voice recording
Build a careful cleanup pass using noise reduction, filtering, dynamics, and level checks.
On this page
Identify the problem before processing
Save a project copy and listen to a passage with quiet speech, louder speech, and a pause. Is the problem a constant hiss, low rumble, a narrow hum, noisy gaps, or an uneven voice level? Each calls for a different tool.
If a stereo recording has the voice on one side only, consider Process → Channel Mixer… → Extract Channel. Preview the correct channel before applying to all audio.
Reduce only the noise you can identify
- For steady hiss, select noise without speech and capture a profile in Noise Reduction….
- Select the voice passage, preview the reduced audio and the residue, and lower the reduction if words sound watery or show up in the residue. Switch Output back to reduced audio before applying.
- For rumble, try a High-Pass Filter with a conservative cutoff. For a narrow hum, inspect Plot Spectrum and use a Notch Filter.
- If only the gaps need quieting, try the Noise Gate and check the ends of quiet words.
These are alternatives, not a checklist to apply in full. Undo anything that reduces intelligibility.
Control levels and check the result
Use the Compressor if loud and quiet phrases differ too much, adjusting threshold and ratio while listening; makeup gain also raises residual noise. Use Normalize for an overall peak target or the Limiter to hold sample peaks.
Compare before and after at the same listening volume. Play a quiet ending and a loud consonant, save the project, and export. Check the export with headphones as well as your usual speaker.