Dear Robert:

Thanks for sharing the experience of using ChatGPT for music creation. I tried to generate/simulate the soundtrack of The Godfather several years ago. (https://franky07724-57962.medium.com/what-i-learned-from-the-godfather-f2398c136372) My approach was based on a note-by-note generation process and you can hear the original piece and the generated/simulated one, but skip the programming part, in my blog. I found that my note-by-note method was not good for two reasons: 1. It presented notes only in a single channel and a single instrument. 2. It presented notes without timing information so that all the notes could only be played at a constant speed. I am wondering if the chord-based approach and the note-based approach share similar problems so that we should let ChatGPT skip chords.

Regards

Frank