marcellieb / musicsync Goto Github PK

Optimize Sample rate and buffer size to allow constant windows and loops

Proper logging

And add more advanced preprocessing

A fixed buffer size is required for the audio to be interpretable in the time domain.
The current seemingly random buffer sizes result in poor detection performance and remove the ability to optimize detection parameters.

It was assumed that flatten denests in an ordering like zip

Add an abstraction layer, that handels

• different polling rates
• combination of offsets
• Animation of onsets
• multiple kinds of lighting configurations

how do I use it D:

• module
• hfc
• spec_flux
• threshold

current stereo to mono conversion is very phase dependent and can lead to inconsistent results depending on the song

Adaptive decay length and rgb envelopes for color fades

There are a few security problems detected by Github

• #54
  • module
  • hfc
  • spec_flux
  • threshold
• #55
  • hue
  • wled
  • serialize
  • console
  • envelope
• #56
• benchmark
• plot

Detect things like high hats, kick and snare drums and melodies separately

This results in a big performance hit

This library allows the use CWT in real-time applications instead of STFT, which allows for higher frequency resolution.

Add weight functions for each detection category. Multiplying the weight with the spectrum in an effort to fit the characteristic frequencies.

• hue
• wled
• serialize
• console
• envelope

Improve the usability of Lightservices and decouple the audio processing module from it

Write onsets to file for comparison between parameters and visualization of precision.

If you stop the application shortly after starting plotting might get stuck consuming 100% of one thread