Audio Tools

Denoise, enhance, and process audio files with AI-powered tools.

Overview

The Audio Tools API provides a suite of AI-powered processing utilities for cleaning up and improving audio quality. Remove background noise, normalize volume levels, trim silence, and target specific loudness standards -- all through simple API calls.

Background noise removal

Audio enhancement

Volume normalization

Silence trimming

Loudness targeting

Multiple formats

Quickstart

Clean up a noisy audio file in just a few lines of code.

from nur import NurClient
client = NurClient()
# Denoise an audio file
result = client.audio.denoise(
    file="recording.mp3",
    intensity="medium"
)
print(f"Clean audio: {result.audio_url}")
print(f"Noise reduced by {result.noise_reduction_db} dB")

Denoise Audio

POST/v1/audio/denoise

Remove background noise from an audio file. The AI model identifies and suppresses unwanted noise while preserving speech clarity and musical content. Supports three intensity levels to balance noise removal with audio fidelity.

Parameter	Type	Description
file	file	Audio file upload. Provide either file or audio_url.
audio_url	string	Public URL of the audio file to denoise.
intensity	string	Noise removal strength: low, medium, or high. Default: medium.

# Denoise from a file upload with high intensity
result = client.audio.denoise(
    file="noisy_interview.wav",
    intensity="high"
)
print(f"Audio URL: {result.audio_url}")
print(f"Duration: {result.duration}s")
print(f"Format: {result.format}")
print(f"Noise reduction: {result.noise_reduction_db} dB")
# Denoise from a URL with low intensity
result = client.audio.denoise(
    audio_url="https://example.com/recording.mp3",
    intensity="low"
)

Response

{
  "audio_url": "https://cdn.nur.ai/audio/proc_x7k9m2v4/denoised.wav",
  "duration": 185.4,
  "format": "wav",
  "original_loudness": -22.3,
  "output_loudness": -21.8,
  "noise_reduction_db": 18.5
}

Enhance Audio

POST/v1/audio/enhance

Enhance an audio file with volume normalization, silence trimming, and loudness targeting. Combine multiple processing steps in a single request for efficient audio post-production.

Parameter	Type	Description
fileREQUIRED	file	Audio file to enhance.
normalize	boolean	Apply volume normalization. Default: true.
remove_silence	boolean	Trim leading, trailing, and excessive internal silence. Default: false.
target_loudness	number	Target integrated loudness in LUFS (e.g., -14 for streaming). Default: -16.

# Enhance for podcast publishing (-14 LUFS for Spotify)
result = client.audio.enhance(
    file="raw_episode.mp3",
    normalize=True,
    remove_silence=True,
    target_loudness=-14
)
print(f"Enhanced audio: {result.audio_url}")
print(f"Duration: {result.duration}s")
print(f"Original loudness: {result.original_loudness} LUFS")
print(f"Output loudness: {result.output_loudness} LUFS")
# Simple normalization only
result = client.audio.enhance(
    file="voice_memo.wav",
    normalize=True,
    remove_silence=False
)

Response

{
  "audio_url": "https://cdn.nur.ai/audio/proc_p3q8r1t6/enhanced.mp3",
  "duration": 1842.7,
  "format": "mp3",
  "original_loudness": -24.1,
  "output_loudness": -14.0,
  "noise_reduction_db": 0
}

Response Objects

Both the Denoise and Enhance endpoints return the same response object structure.

AudioResult

Field	Type	Description
audio_url	string	URL to the processed audio file.
duration	number	Duration of the output audio in seconds.
format	string	Output audio format (mp3, wav, flac, ogg).
original_loudness	number	Integrated loudness of the input in LUFS.
output_loudness	number	Integrated loudness of the output in LUFS.
noise_reduction_db	number	Amount of noise removed in decibels. 0 if denoise was not applied.

Best Practices

Denoise before enhancing

Run the denoise endpoint first, then pass the cleaned audio to enhance. This produces better normalization and loudness results since the algorithm is not influenced by background noise.

Match loudness to your platform

Use -14 LUFS for Spotify and YouTube, -16 LUFS for Apple Podcasts, and -24 LUFS for broadcast television. Setting the correct target_loudness avoids re-encoding by the platform.

Use low intensity for music, high for speech

Music content is more sensitive to aggressive denoising, which can remove subtle harmonics. Use low or medium intensity for music and reserve high for speech recordings.

Chain tools in a pipeline

Combine denoise and enhance in sequence to build a complete audio post-production pipeline. Automate this with webhooks or SDK callbacks for batch processing of large libraries.