ComfyUI Node Runs on cloud
Audio De-Esser
Turn down harsh s, sh and t sounds in a voice only when they flare, leaving the rest untouched, such as smoothing a bright text-to-speech take. Listen mode plays the band being caught, to tune the frequency.
Audio De-Esser
- audio
- audio
- max_reduction
◄frequency6000►
◄threshold_db-30.0►
◄range_db8.0►
◄modesplit band►
◄output_db0.0►
CategoryWAS Suite/Audio/Dynamics
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| audio | AUDIO | The voice to de-ess. | |
| frequency | FLOAT | 60002000–16000 | Hertz the sibilance band starts at. `5000` = deeper voices; `6000` = most voices; `8000` = bright female voices and TTS. |
| threshold_db | FLOAT | -30.0-60–0 | Level in dBFS of the band above which it is turned down. Lower catches more: `-40` = heavy; `-20` = only the worst. |
| range_db | FLOAT | 8.00–24 | Most dB the band is turned down. `4` = subtle; `8` = typical; `16` = lisping. |
| mode | COMBO | split band | `split band` = only the high band goes down; `wideband` = the whole voice dips, more natural on light sibilance; `listen` = output the band caught. |
| output_db | FLOAT | 0.0-24–24 | dB applied to the result. `0` = unchanged. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| audio | AUDIO | The de-essed voice, or the band caught in listen mode. |
| max_reduction | FLOAT | Most dB the band was turned down on the first clip. |