VeedCleanAudioNode - ComfyUI Built-in Node Documentation - ComfyUI

Remove background noise from speech with VEED. It is built for speech in any language: noise under the voice is cleaned, pauses are cleaned completely, and music and sound effects are removed rather than preserved. The output is 48 kHz mono audio.

Inputs

Parameter	Description	Data Type	Required	Range
`audio`	Speech recording, shorter than 30 minutes.	AUDIO	Yes	Up to 1799 seconds
`strength`	How much noise is removed under speech. Lower keeps more room tone behind the voice; pauses between words are always fully cleaned, and values above the default change little. (default: 0.874)	FLOAT	Yes	0.0 to 1.0 (step 0.001)
`loudness`	Whether to normalize the output loudness. `"normalize"` sets the output to `target_lufs`; `"keep input level"` leaves the level as it is. Selecting an option reveals its parameters. (default: `"normalize"`)	DYNAMIC_COMBO	Yes	`"normalize"` `"keep input level"`
`target_lufs`	Integrated loudness of the output in LUFS. Peaks are capped at -1.1 dBTP, so loud targets stop early (about -15 LUFS for typical speech). Available when `loudness` is `"normalize"` (default: -19.0).	FLOAT	Yes	-40.0 to -8.0 (step 0.1)

The node stops with an error when the recording is longer than 1799 seconds.

Outputs

Output Name	Description	Data Type
`AUDIO`	The cleaned speech as 48 kHz mono audio.	AUDIO

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): c85949385109a5d320e52619e9df17c245b19f8f16cc1a0b69a37ae2e7bb5b6d