I'm making an automatic music transcription system for singing so that you can convert your voice to midi. To train the system, I need WAVs of singing that are perfectly synched to midi files.
Does anyone have any suggestions of the best way to get this training set?
One thought was to take a midi file and then use it to produce vocalese. However, I don't know where I might find good samples. This obviously isn't as good as true singing, but could at least serve as a simple baseline.
Another option would be to find some pre-recorded monophonic singing for which midi is available and then using dynamic time warping to sync them. Again, I would need to find a source for this.