Each "circle" is a specific frequency. It becomes larger and brighter depending on frequency's magnitude.
Every frame a FFT is performed on a set of samples, with a blackman window. Then a threshold is applied to hide low-magnitude-frequencies (Magnitude Threshold dial).
After that, an array of each frequencies' magnitude is sent to the graphic card, computing the image.

Since a FFT is performed each frames, it is possible to react to a mic's input in real-time with a very small lag.

On the video, you can see a right panel with parameters.

• Magnitude Threshold : Defines a threshold to cut low magnitude-frequencies
• Square count : Define the amount of frequencies taken, resulting to a more or less sharp visual
• Red/Green/Blue groups : Defines how colors react to magnitude. "R" means it takes only right-speaker frequencies, "L" for left-speaker and "I" means it scales on frequency ID

I've used Qt for UI and OpenGL + GLSL to display.
To perform FFT, I've used the mathematical library LibFFTW3