+Bharat Mallapur No Bharat, the Speech Engine and Facial Recog work independently of each other. The Speech recog processes the mic input in the background, Although I must say, that's an excellent suggestion you've made- to initiate speech recog only after face detection.

+Saransh Kejriwal yeah, that could actually help, since you could load the acoustic profile of the particular speaker in question. But I was talking more about using visemes to increase accuracy of phoneme detection...

+Bharat Mallapur That would actually help, provided that only one sample set, ie one person, will be operational. Viseme processing would depend on several factors like facial structure, lip size, linguistic accent etc., so we'd require different training models for different people.

Thanks a lot Shroom. Good luck with your setup

+Huseyin Guler github.com/SaranshKejriwal/Harold_Finch