I have custom trained a model by building a CNN architecture to detect different sounds produced by a machine for condition monitoring. Now I want to run this model in real-time so it can take in audio input from a mic and pre-process the data, pass it through the model (.h5) file and predict which label the audio belongs to.

I do not have much experience in programming but whatever code I wrote gave me shape input error and I am not able to resolve it, neither do I know whether what I have written is the right approach or not.