For integration with DAWs, I'd recommend using the SFZ v2 format, which is an open standard, uses a plaintext file for the metadata to specify how WAVs are mapped to the keyboard, has many extra features like release samples and the ability to repitch as needed to avoid the sox conversion, and is widely supported with VST and AU plugins.
For the wind-up sound and whirring noise, you can map those to keys outside of the range of the music box (C2 and C#2 for example).
http://www.sfzformat.com/index.php?title=Main_Page