[Update] node-red-contrib-deepspeech-stt v0.4.0 (added streaming inference support)


I just published version 0.4.0 of node-red-contrib-deepspeech-stt:

This version brings the major addition of support for the streaming api in form of a deepspeech-stream node.
This allows the live transcription of audio while it is streaming in addition to the previously available wav file transcription.
This for example allows on hardware like a pi 4 to have the inference result ready nearly as soon as the audio stream finished.
This should make the user experience when using the node in projects such as diy smart speakers much faster if used in this streaming mode.

As always i am happy about any bug reports or feature requests on the nodes github repository:

Best regards Johannes


Very interesting node.

  1. Does it support other languages than English (eg Dutch) ?

  2. What is the cpu consumption on a raspberry pi 3 (or 4) when running this continuously (eg for hot word detection) ?

In the documentation I link a thread on the Mozilla forum where there is a collection of other available languages. I don't know about Dutch bit there is languages like Chinese, German, Italian and some others.

It's not suited for Hot word detection. Depending on the model and beam width transcription performance on a pi 4 is roughly real time. If you are looking for an easy to use hot word / Kew word spotting system have a look at another node of mine:

Which runs well on even a pi 3a+.


1 Like