Noise Suppressor · API

Remove noise

Clean one audio file and get the cleaned audio back.

POSThttps://internal.audiclabs.com/v1/noise-suppressor/denoise

Request

Send multipart/form-data. Costs 1 credit per call. Needs the noise-suppressor:denoise scope.

filefilerequired

The audio to clean. Any common format and sample rate, up to 100 MB.

modelstringoptional

Model id from GET /models. Default audic-2.

attenuation_limit_dbnumberoptional

Maximum noise reduction in dB, 0 to 100. 0 (default) removes as much noise as possible.

output_formatwav | flacoptional

Encoding of the returned audio. Default wav.

keep_input_sample_ratebooleanoptional

Return audio at your file's sample rate (true, default) or the model's native rate.

responseaudio | jsonoptional

audio (default) returns the file; json returns {filename, stats, audio_base64}.

Example

curl https://internal.audiclabs.com/v1/noise-suppressor/denoise \
  -H "Authorization: Bearer $AUDICLABS_API_KEY" \
  -F file=@noisy.wav \
  -F model=audic-2 \
  -F output_format=wav \
  -o clean.wav

Response

200 with the cleaned audio (audio/wav or audio/flac) as an attachment named <name>_denoised.<format>. Processing stats are in the X-NS-Stats header:

{
  "model": "audic-2",
  "model_sample_rate": 16000,
  "input_sample_rate": 16000,
  "output_sample_rate": 16000,
  "duration_s": 9.769,
  "processing_s": 3.725,
  "rtf": 0.381,
  "attenuation_limit_db": 0.0
}

Errors use the standard format, see Errors. An unknown model returns 400.

Guide

Remove noise (bulk)