How to fix harsh s sounds after cleaning up a voice recording
Ask Vizard Agent for a clean, clear voice and say it has to sound natural on a phone speaker. It measures the s sounds against the main speech range before and after processing. If the clean-up made them stand out, it moves the de-esser after the compressor and works on the hissing band alone until they sit lower than in the raw recording.
What do you type into Vizard Agent?
Say what the voice is for and where people will hear it. Hissing consonants are easy to notice on a phone speaker, and studio sound to most people means louder, denser and brighter, which is exactly what makes s sounds sharper. Naming the device tells Vizard Agent to keep them in check.
Make the sound studio quality with strong noise reduction. Light compression and de-essing, and it has to be clear on a phone speaker without headphones.
What does Vizard Agent actually do?
It checks whether its own processing caused the problem. A presence boost and heavy compression both make the top end more prominent, and a compressor placed after the de-esser raises the s sounds again. In the session behind this article it:
- Built the clean-up chain, then kept hearing the s sounds flagged as harsh, even after turning the de-esser up twice.
- Stopped relying on listening and measured the peaks of the hissing range, 5 to 8 kHz, against the main speech range, 1 to 3 kHz, in the raw and processed voice.
- Found the s sounds sat 10.3 dB below the speech in the raw recording but only 5.5 dB below after processing, so the clean-up had made them about 5 dB more prominent.
- Traced it to the main compressor running after the de-esser, moved de-essing after it, and limited it to the hissing band, with fast, hard compression there and a small fixed cut.
- Re-measured the gap at 11.5 dB, then 13.3 dB after a final pass, with the voice at −14.4 LUFS.
When does this not work well?
When the harshness is in the recording itself, such as a mic held too close to the mouth or a phone mic that distorts on every s. De-essing can soften those peaks but cannot rebuild a distorted sound, and pushing it too hard makes the speaker sound like they have a lisp.
Common questions
Why did the clean-up make it worse? Compression after de-essing lifts them again. Vizard Agent reorders the chain.
How does it know they are fixed? Vizard Agent measures them against the speech, not just by ear.
Will the voice sound dull? No. Vizard Agent works on the hissing band only.