Use this demo to test out the quality of realtime denoising for your input audio, specified chunk size and the output sampling frequency. The demo runs the script provided in the stream from a file section of the quick start guide.

The demo does not stream audio from you to our servers, so you simply provide a .wav file, but it will chunk your audio file, run denoising in realtime and stream back the resulting audio. You can use this to test out the quality of audio denoised in realtime.

If you would like to deploy this on your own servers in true realtime, this is available today. Please contact us to get started.

The arguments below are as follows:

  • Audio File Upload a .wav audio file which you would like to remove background voices from or select from our provided examples.
  • Chunk Size in ms Provide the time in ms of the chunks of audio clip that you would like to send to the denoiser. The file that you upload is split into chunks this size and sent to the denoiser to emulate realtime streaming. As our model currently runs inference at ~63ms, the minimum chunk size is 65ms.
  • Sampling Frequency of Denoised Audio (Hz) Our model downsamples your input to 8000Hz, does processing at this rate and then upsamples the output back to the value you provide here. We are working on increasing the frame rate that the model can process to remove the need to up and downsample.