Then you should use the mincer opcode. Check out the examples in the manual and if you have any problems post some code here. I can’t really write much code at the moment because I chopped off my hand two weeks ago, but the voice to text function works pretty good so I can discuss and help you that way.
This is a piece of cake for Cabbage, easy straight out of the box.
No problem building a complex UI with cabbage as long as you’re happy to prepare good assets. These plugins are a great example of what can be done in this way:
https://gbsoundlab.com/
p.s. Packaging the samples might be an issue, currently there is no way to encrypt them or prevent users from opening up the bundle and accessing the samples directly. Cabbage doesn’t have a proprietary sample format. This might be a problem for you?