Silverhead, you are correct. This would take some effort. In my mind the coding would be minor compared to the planning and testing.
We would need to break all conversions into three options:
As Is: This is where the conversions match as is. (This sounds like your Phase 1)
Simple: This is where there isn't a match but decisions can be made by the program. (This sounds like Phase 2)
Complex: This is one that would require a person to decide what changes to make. This could be done with or without the program making suggestions.
As the models are in JSON, it should be easy to create objects, modify them, then spit the JSON out for the right output. Except for UI and decision making, a lot of the code should be straightforward.
However, as you can see by the document I attached, discovery is a big animal. Determining what models exist, how they map to other models, how branches work between models, when users need to be involved in decision making, etc. would be a much bigger piece. Then, testing these mappings on various devices would be a big animal.
The plus side is that most anyone can do most of the work here, with just a little bit of direction. Coding would be possibly 25-30% of the total work.
The thing I find most intriguing is that CustomForge has done the work of decoding a ton of tones for Rocksmith (Just under 6000). Tapping into that resource to pull tones for Pod Go and Helix seems like a huge win.
You are correct, Silverhead: this will be a big undertaking. It would provide access to many tones and convert between models. The good news is anyone who is interested can contribute if they are willing to read JSON and/or spreadsheets and document them. if that sounds valuable to anyone, just reply. If not, no worries.