Is there a streaming api (realtime)

Apols as a 1st look @ Speechbrain and just thought of trying a stereo sounddevice input and confused.

The examples are for files? Specifically the multi_mic is it possible to copy from the sounddevice numpy frame into stft?

Hey, as listed in the roadmap streaming isn’t available yet. It wouldn’t be super hard to add, but we do not have the resources yet to work on that. If you end up doing something, please share it with us ! Even if it is super minimal !


My code is far to hacky but have been trying to give it a go by not using Torch because of the MKL probs (Arm64).

I gave it a go and ground to halt but will mention what I was trying and failing.
I have been trying to get the interest of others who are a bit more Pythonic than I, as hopefully the colab is a great tutorial.

I will shout and post if anything comes of it.

Remove torch from SpeechBrain is going to be pretty hard :stuck_out_tongue:

Torch methods from STFT… Where there is no need for non AI beamformers to use Torch Math.
Any math will do and I prob could of got further with Arm64 but hey.