Fixing Sibilance and Plosives in Voice-Clone Samples
Harsh S sounds and popping P/B hits in the clone usually started in the recording. Mic angle, distance, and a simple listen-back beat software EQ.
The clone exaggerates what your mic already captured
If your samples have sharp, whistling S sounds (sibilance) or a gust of air on P and B (plosives), the cover will have them on every lyric that uses those sounds — which, in English pop, is most of them. People hear that as "cheap AI." It is usually cheap placement.
This is a recording problem, not a reason to jump to a new song. Clean samples is the general checklist; this page is the two consonants that cause the most complaints.
Plosives: stop aiming the mic at your breath
A plosive is a burst of air hitting the capsule. Phone mics are tiny and sit right in that blast if you hold the device like a walkie-talkie.
- Move off-axis: point the mic at your chin or slightly to the side, not straight at your teeth
- Keep the hand's width distance from home mic technique
- Say the prompt at normal volume; do not over-enunciate P sounds like a pronunciation drill
- If one take pops, redo that take only
A foam windscreen on a USB mic helps. A finger over an iPhone mic does not — it just muffles everything.
Sibilance: harsh S, not "clear diction"
Some voices are naturally sibilant on cheap capsules. You can still make it usable:
- Do not hiss the S to be "clear." Diction means understandable words, not weaponized consonants
- A slightly darker room voice (less smile, less air through the teeth) often records cleaner than a presenter voice
- Avoid "voice enhance" and bright EQ before upload — they add the exact 6–8 kHz spike you are trying to kill
- If every sample is harsh, the mic is too close. Back up an inch before you buy anything
When the song makes it worse
Bright, dry pop mixes (Espresso, a lot of Sabrina Carpenter) will show sibilance more than a dark, wet ballad. If the clone is only ugly on one track, try a warmer source before you throw out a good model. If it is ugly on every track, fix the samples.
Open the studio and listen to your samples on headphones for pops and hisses before you train — that listen is the whole technique.