Realtime detection of sound

[giusarno]

Hello,
all the examples I have seen attempts to detect sound using the full length of the sound sample.
I would like to implement a system that can detect the sound in Realtime therefore it is important to monitor the audio sample in real-time and generate spectrum images in real-time.
What would be the best approach to do this ?
One option is to listen to the audio and split it in chunks of fixed length. This will mean to inject each chunk into the model until the sound effect is detected. Would this approach work with this model ? is the model capable to recognize sound for smaller audio samples ?

Regards.