Start with the problem, not the tool
Audio repair works best when you name the defect first. Was the recording made in a room with a fridge humming? Is there a mouth click every few seconds? Did the singer clip the preamp? iZotope RX 10 Elements is the entry-level edition of a well-known repair suite, and it focuses on a short list of common fixes. Below is a symptom-led guide you can follow whichever repair tool you use; check the module names in your own copy, since editions differ.
Symptom to fix table
| What you hear | Likely cause | Repair type |
|---|---|---|
| Constant hiss or room tone | Noisy preamp, air conditioning | Broadband noise reduction |
| Steady buzz at 50 or 60 Hz plus harmonics | Ground loop, mains hum | De-hum |
| Short ticks and pops | Mouth noise, digital glitches, vinyl | De-click |
| Harsh flat-topped peaks | Input clipping | De-clip |
| Boomy, uneven voice | Room reflections, proximity | EQ, not repair |
Walkthrough: rescuing a phone-mic vocal
Scenario: a spoken demo recorded in a bedroom, with a faint fan hiss and a 60 Hz hum.
1. Work on a copy
Duplicate the audio file. Repair is destructive, so keep the original untouched.
2. Fix hum first
- Zoom into a section with no speech and find the low line in the spectrogram.
- Set the base frequency to match (50 or 60 Hz, depending on your region).
- Start with a few harmonics and only add more if you still hear buzz.
3. Reduce broadband noise
- Select a second or two of pure noise so the tool can learn it.
- Begin with a modest reduction, roughly 6 to 10 dB.
- Listen to what you remove. If you hear syllables in the difference signal, back off.
4. Click and clip cleanup
Run de-click at low sensitivity, then check the result by soloing what was removed. Only use de-clip if the waveform has obviously flat tops.
Rules that prevent artefacts
- Several light passes beat one heavy pass.
- If the voice starts to sound watery or "swirly", you have gone too far.
- Noise reduction goes before EQ, compression and reverb, never after.
- Do not repair music with the same settings you use on speech.
Quick A/B procedure
- Bounce the original and the repaired file.
- Level match them.
- Listen on headphones at normal volume, then turn it up loudly; artefacts show at high volume.
- Pick the version with less noise and natural consonants.
Prevention beats repair
Most fixes disappear if you record smarter: move away from the fridge, turn off fans, use a pop filter, and keep input peaks around -12 dBFS. Our gain staging guide shows where to set input levels so you avoid clipping in the first place.
After the repair
Once the noise is gone, treat the voice normally. A gentle compressor helps even it out, and the compressor tutorial covers threshold, ratio and release. To try more free processing tools, browse the free VST list.
Availability note
Free offers for RX editions have been time-limited in the past. Check the current price, licence and included modules on iZotope's official website, and download only from there.
Repairing music versus speech
Speech tolerates strong noise reduction because it is sparse and has short gaps. Music does not. For a field recording of an acoustic guitar, keep reduction near 3 to 6 dB, and always check the sustain tails, where over-processing shows first as a strange grainy fade.
Working order for a podcast episode
- Trim silences and breaths manually where they are obvious.
- De-hum the whole file if a tone is constant.
- De-noise with a learned profile from a quiet spot.
- De-click mouth noises on the loudest sections only.
- Only then apply EQ, compression and loudness normalisation.
Reading the spectrogram
Time runs left to right and frequency bottom to top. Steady noise looks like a horizontal band, a click looks like a vertical line, and a hum shows as evenly spaced horizontal stripes. Learning to spot these three shapes makes choosing the right tool automatic.
Second scenario: a clipped guitar-amp room recording
You recorded a live rehearsal on a handheld recorder and the loudest chorus is distorted. Zoom in on the waveform: peaks flattened at the top confirm clipping. Apply de-clip with a moderate setting, listen, and if it sounds better but still rough, run a second gentle pass rather than raising the amount. Follow with a small cut around 3 kHz if harshness remains. Do not expect to restore original dynamics; aim for less fizz, not perfection.
Troubleshooting
- Voice sounds underwater: reduction is too high or the noise sample included speech.
- Hum comes back after export: a harmonic was missed; add another and re-check the spectrogram.
- Breaths vanish: lower reduction; natural breaths make speech believable.
- Different clips do not match: apply similar settings to all and add a bit of shared room tone.
FAQ
Can repair fix a badly clipped recording?
It can reduce harshness, but heavy clipping cannot be fully reversed. Re-record if you can.
How much noise reduction is too much?
When consonants like "s" and "t" lose their edge or you hear a warbling tone in the gaps, reduce the amount.
Should I repair each clip or the whole track?
Repair each source separately, because different clips have different noise profiles.
See RX Elements at iZotope
