NGC | Catalog
Welcome Guest
CatalogContainersMaxine Audio Effects SDK version 1.0

Maxine Audio Effects SDK version 1.0

For pull tags and more information, please view on a desktop device.
Logo for Maxine Audio Effects SDK version 1.0


Maxine Linux Audio Effects SDK Public container version 1.0



Latest Tag



July 1, 2022

Compressed Size

12.94 GB

Multinode Support


Multi-Arch Support


1.0.5 (Latest) Scan Results

Linux / amd64

Audio Effects SDK

The AFX SDK provides features to reduce unwanted audio noise. It comes with two features, Noise Reduction (NR) and Room Echo Removal (RER). To provide a pleasant experience, NR removes several common background noises using state-of-the-art AI models while preserving the speaker's natural voice. RER removes reverberations from audio using state-of-the-art AI models, restoring clarity of a speaker's voice. There is also a model that combines both features into one.


There are three main prerequisites for AFX SDK's containers:

  • NVIDIA Drivers NVIDIA Drivers (465.19.01+) for recommended for Maxine) are required to use NVIDIA GPUs here. Note: For older drivers please review CUDA Backward Compatibilty(experience may or may not be optimal).
  • Docker (19.03+) and the latest version of NVIDIA-Docker
  • NGC API Key for logging to NVIDIA's registry. Details are available here.

Running the container

Use the following commands to run the container.

docker run --gpus all -it --rm -v :

and if using depricated nvidia-docker, based on a version of docker prior to 19.03, use:

nvidia-docker run -it --rm -v /tmp/.X11-unix:/tmp/.X11-unix -w

Documentation & resources

There are a plethora of resources you can access like our Programming Guide, Getting Started Guide, Devloper Blogs and API Videos. All these resoureces are accessible via our Getting Started Page.

The AFX SDK's programming guide can be found in the container at /usr/local/AudioFX/docs and here. This guide contains:

  • Supplementary feature information and information for selecting the most suitable configurations for specific use-cases
  • Detailed AFX API descriptions with best practices and examples
  • Information for using sample applications
  • Bare-metal installation requirements

The container houses sample applications for feature demonstrations as well as an implementation example. Check the "Running a sample application" section for more details.

The sample applications provide an example for using Maxine-specific APIs. Some familiarity with CUDA is required. To brush up on CUDA basics, check out the tutorials, the documentation, and the best practices guide.

Running a sample application

AFX SDK's sample applications can be found in /usr/local/AudioFX/samples/ which contains effects_demo and effects_delayed_streams_demo. The following example demonstrates how to use the effects_demo application.

./effects_demo -c turing_denoise16k_1_cfg.txt

The above command processes one 16kHz audio file on a Turing gpu. Let's take a closer look at the used config file.

effect denoiser
sample_rate 16000
model ../../models/turing/denoiser_16k.trtpkg
real_time 0
intensity_ratio 1.0
input_wav_list ../input_files/denoiser/16k/Fan_16k.wav
output_wav_list out_16k.wav

The user can select the effect with effect (denoiser, dereverb, dereverb_denoiser), specify the I/O files with input_wav_list and output_wav_list, choose the sample rate with sample_rate(16 kHz or 48 kHz). Please explore Section 2 of the programming guide for more details.


  • Naming convention for models has been changed (see Section 3.2 (Setting the Parameters of an Audio Effect) in Programming Guide for details).
  • The SDK supports MIG on NVIDIA DGX A100
  • Improved the quality of the Noise Suppression Effect.
  • The maximum batch sizes supported by some effects have changed.


The Audio Effects SDK license can be found within the container and on our Getting Started Page

Ethical AI

NVIDIA's platforms and application frameworks enable developers to build a wide array of AI applications. Please, consider potential algorithmic bias when choosing or creating the models being deployed. Work with the model's developer to ensure:

  • The model meets the requirements for the relevant industry and use case
  • The necessary instruction and documentation are provided to understand error rates, confidence intervals, and results
  • The model is being used under the conditions and in the manner intended.