MariHacks2k20

• 

  Our logo

• 

  Screen capture of the results in audactiy

MariHacks2k20

Melodia: a auto vocal accompaniment software

This is the repl.it we coded it on. https://repl.it/@NicholasFicara/MariHacks

What inspired us? When a vocalist wants to create an instrument track, they either need to create it themselves, which requires knowledge of instruments or pay someone to do it for them. Our solution is easy to use and effective at creating an instrument track for any vocal recording. We are all pianists so we felt like we were up to the challenge of making this app.

What we learnt. Audio itself is extremely complex. We learned how to properly use several audio libraries in Python3 and we got a deeper understanding of the fundementals of audio. We also learnt the importance of good test data. At first, we were working with awful audio. We realised that it was one of the biggest issues.

How we built it. We used Python3 to build the application along with the parselmouth library. We passed the audio through a pitch extraction algorithm, then ran feature extraction. Next, we ran a custom grouping and timing algorithm for the parsed features. We then used several music theories to populate a midi file with the proper notes. We used Audacity to play the midi file and sync it with the original vocal track.

Challenges we faced. We ran into countless issues extracting pitch and features from audio. We also ran into many issues exporting the midi to audio files. Making the music sound good was probably the biggest challenge. We are all happy with the end result though.

Built With

• python