TabifAI creates high quality, accurate music transcription through a generative, polyphonic neural network.

You can either upload an existing audio file (.wav, .mp3, etc.) or record live directly in the application. Then model will extract single notes, triads, and more advanced chords which it will use to generate a piano transcription of the song or a midi transcription.

TabifAI's transcription service has a place in every musicians' toolkit:

  • experienced improvisationalists often don't remember exactly what notes they played in a particular solo so feeding the song live into tablature would allow them instant access to all the content they produce.
  • beginner musicians can learn music theory in a much more intuitive way by visualizing the notes they play on the screen
  • tabs no longer need to be generated manually
  • midi files can be transferred to any and all production toolkits for an improved workflow and a more seamless production experience

TabifAI is entirely web based. No need for clunky servers because even the ml audio preprocessing is done directly in the browser using Tensorflow.

Our biggest challenge was training the ml to be polyphonic - meaning it can recognize multiple pitches and tones playing simultaneously.

Built With

Share this project:

Updates