PianoTranscription

Generates midi tracks on Piano from music using Machine Learning. This is the macOS executable and app of Piano transcription

Install

Please use the App Store Version, which uses CoreML instead of PyTorch.

Files and source code could be found in releases.

Note: If mac says the app was damaged / unknown developer, please go to System Preferences > Security & Privacy > General, and click Open Anyway. Show Details.

Interface

Written with SwiftUI.

Example

• Variations on the Kanon (George Winston, December, by Pachelbel)
• Ballade No. 1 in G minor, Op. 23 (四月は君の嘘, 四月は君の嘘 僕と君との音楽帳, by Chopin)

Portable Executable File

All the files are complied from Piano transcription.

This is a portable executable file, hence you do not need to install anything.

Arguments:

--audio_path: The path for the source.
--output_midi_path: The path for the destination.
--checkpoint_path: The path for checkpoint file

the checkpoint file can be obtained from Piano transcription

Additional Arguments:

These arguments are optional

--cuda: A boolean value determining whether gpu should be used.
--onsetThreshold
--offsetThreshold
--frameThreshold
--padelOffsetThreshold
--batchSize

Example:

PianoTranscription --audio_path='music.m4a' --output_midi_path='file.midi' --checkpoint_path='checkpoint.pth'

If you want to build it yourself:

install pyinstaller > In your terminal,

pyinstaller example.py --collect-all piano_transcription_inference --collect-all torch --collect-all librosa --collect-all sklearn --collect-all scipy --collect-all audioread --collect-all numpy --collect-all mido

Note: You would find the files in Build Files helpful

Credits

• Piano transcription
• Swift Tagger