Recognition accuracy crucially depends on the musical nature of the audio record at hands as well as on the recognition parameters you set. Choosing optimal parameter set allows to make MIDI output to sound as close as possible to the initial music. In particular, the optimal choice minimizes the number of spurious notes that may appear.
Keep in mind that transcribing a piece of music played with different instruments vary in difficulty. Piano or clear guitar music is relatively easy to transcribe, and the MIDI output may sound very close to the original. Winds are usually transcribed with lesser success, whereas bow instruments are very hard to transcribed automatically at all. Also an attempt to transcribe human singing, especially when it comes with lyrics, usually meets considerable difficulties.
Main plug-in's page contains three sliders. They control polyphony, sensitivity and output velocity. For a start, we recommend to set all of the sliders to their maximum. Recognition with these settings will certainly result in a number of waste notes, but on the positive side you will get an idea of the pitch range for the melody you work on. Set up the estimated range using Equalizer. In order to improve the quality of recognition, you may need to reduce somewhat the bar height corresponding to both lowest and highest frequencies of the range.
After setting up Equalizer, go back to the main page and gradually lower sensitive value, until most of the waste notes are gone. Stop just above the level where the useful notes begin to disappear. Then, decrease the polyphony parameter to further reduce remaining waste notes.
It is recommended to adjust settings while the audio file is being played back. It makes it possible to hear the immediate result of the adjustment and see it on the spectrogram. Once settings are chosen, proceed with recording of the resultant MIDI sequence to the MIDI-track to a MIDI file with a built-in MIDI recorder (you can choose output method on options page).
Note: Notes played back during recognition process may last shorter or longer than those recorded to MIDI track of file. The actual note duration can only be measured after the note is played. Therefore, durations of the recorded notes are known with greater precision.