This one uses a simple, command based speech recognition system running on the browser - thanks to webassembly! The model is trained using Kaldi and PyTorch.