Noise Cancellation

SDK Documentation

C++ SDK documentation here.
Python SDK documentation (will be available soon)

Example

The technical integration of Krisp Audio SDK should have correct sequence of functions invocation. The following example shows how to use the exposed functions.

try
{
    globalInit(L"");

    ModelInfo ncModelInfo;
    ncModelInfo.path = weightFilePath; // .kef file

    NcSessionConfig ncCfg =
    {
        inRate,
        frameDurationMillis,
        outRate,
        &ncModelInfo,
        withStats,
        nullptr // Ringtone model cfg for inbound stream case
    };

    std::shared_ptr<Nc<SamplingFormat>> ncSession = Nc<SamplingFormat>::create(ncCfg);

    while(frameAvailable)
    {
        ncSession->process(inFrame, inFrameSize, outFrame, outFrameSize,
                           noiseSuppressionLevel, withStats ? &perFrameStats : nullptr);
    }

    // ncSession is a shared_ptr. Free pointer before globalDestroy()
    ncSession.reset();
    globalDestroy();
}
catch(std::exception& ex)
{
  // Handle exception
}

import krisp_audio

# initialize Krisp SDK global instance
krisp_audio.globalInit("")

# Create Noise Cleaner with the specified configuration
nc_cfg = krisp_audio.NcSessionConfig()
ncFloat = krisp_audio.NcFloat.create(nc_cfg)

# Noise Cleaner frame by frame processing of the given audio stream
for i in range(0, 1000) # frame count
    processed_frame = ncFloat.process(frame, self.suppression_level)

# Free the Krisp SDK global instance
ncFloat = None
krisp_audio.globalDestroy()

Integration

Below are the high level steps required for technical integration:

SDK initialization	the SDK library should be loaded and initialized before being used. The SDK occupies resources that should be released once the SDK is no longer needed.
Model initialization	Krisp SDK ships with NC and BVC models. Each model has specific requirements for the device and the sampling rate. Each model should be initialized and loaded into the memory before it can be used.
Session creation	In order to use Krisp Audio SDK to process the sound stream a Session object should be created. The Session creation is coupled to the AI model which should be specified during the session creation process. The session object also requires the specification of the sampling rate and frame duration.
Frame processing	Frame processing is done using the Session object which works only for the given sampling rate for the given frame duration using the loaded AI model.
Releasing the resources used by the session	the resources occupied by the session should be released once the session is no longer used.
Unloading the SDK	the resources loaded by the Krisp Audio SDK should be released once the SDK is no longer needed

ℹ️ Multiple sessions can be created and used at the same time
ℹ️ Each session can be processed in a different thread on another audio stream
ℹ️ A dummy Session can preload model into memory, while a separate Session could handle the actual stream processing. This approach may be helpful to reduce latency before processing begins, as the second Session will utilize the cached model data preloaded by the first Session and will be created much faster than the first Session.
❗The SDK should be unloaded only after the release of all sessions

Linux Integration

Managing Multithreaded Behavior

By default, the SDK operates in multi-threaded mode to optimize performance and support a higher number of concurrent sessions on a given CPU. The actual number of threads is managed dynamically by the OpenBLAS runtime, typically allowing more sessions to run than the number of available CPU cores.

If your application already processes Krisp Noise Cancellation (NC) in multiple threads, it is recommended to configure the SDK to run in single-threaded mode to prevent excessive thread contention and optimize overall system performance:

export OPENBLAS_NUM_THREADS=1