This program implements a real-time analyzer for audio spectrogram, which is a 2-D time-frequency plot of audio signal. Using the abundant information carried in the spectrogram, it is possible to invent some sound-driven applications, especially for the sound-controlled games. The implementation is based on Python 3 with imports of pygame and pyaudio, so the dependency is rather light compared with some of the other similar implementations found in the internet. A real-time speech recognizer is also a possible extension of this program. I am going to implement a speech recognizer based on this program in the next few weeks and hope it will be done before the conference date in May.