some reviews says the melody extractor generates midi in random, it maybe the previous models limitation, while we do never generate midi randomly, all midi data is from AIs prediction.
Train a great model is not very easy, if the model is not good enough, the data from models prediction may make people feel it like generated randomly.
Now, we had taken a new neural network method, and trained a new good model, thanks for TensorFlow making it much quicker and easier.
You could extract the melody from any music or even voice by using our melody extractor.
Singing a song, then extract the melody, export the midi file, it will even make your music creation more easier.
Though we try to use the newest and the best model to extract melody from music, AI could not duplicate the songs original staff melody anyway, AI can only keep a probably notes for midi, and some notes maybe wrong.
We do not promise you get the perfect melody for each song, maybe one songs melody is very good, while another songs melody is a little bad. The results depends on the songs melody itself, if the songs melody is similar with the AI models trained data, the result will be very good, if not, the result will be a little bad.
Anyway, its worth you have a try, melody extractor will help you get pure melody from any music more simple.
Notes: theres two common reason for this apps crash, one is the music file too large, maybe too long, this will make memory out of usage, we should not consider it as a bug, the mobile device has limitation on memory itself, we could use audio trim tools to clip the audio, try to make the audio file smaller size, then it should be ok for processing to extract the melody. And another reason is midi player crash, as we use the piano file embedded into app, some notes may be lack of, so we should not rely on this app to play midi file, we should consider to export the midi file to computer, and let the midi software on computer to play the midi file.
Surely, for most of music files, these two issues should not happen, as most of music files is not very long, and the embedded piano file should be enough to play most midi file.