My idea to make WhatTheJack more usable would be to add two things.
- A grammar of the jack/submod types (It may already have this) and assigning types to all of the jacks in any mod. The more properties you can have in the grammar about the submods' functions, the more useful.
-
A script or whatever to compile the wiring patterns from uploaded patches* into an xml or whatever. (* To the extent you can get that data. Using the grammar, they could be from any system, not only VCV.)
-
A script that uses any number of ML algorithms on the data to create a predictive model of wiring patterns. (I looked up ones for autocomplete and they looked in the neighborhood. It takes all the Ins and Outs in any given rack and generates a weight table of probability.)
Then for whatever modules are on the rack, the jack placement would be weighted according to the likelihood a connection of A type jacks into a connection of B type as used in actual patches. Then you could have a knob where 0 is precisely the most likely, and 1 is what you have now (completely randomized), so you could get the level of randomness you want. If there's not enough data, then I think you'll want to generalize the grammar more; so it could make a model where roughly submods of type-A generally more often jack into submods of type-B, and add that to the weighting.
My idea to make WhatTheJack more usable would be to add two things.
A script or whatever to compile the wiring patterns from uploaded patches* into an xml or whatever. (* To the extent you can get that data. Using the grammar, they could be from any system, not only VCV.)
A script that uses any number of ML algorithms on the data to create a predictive model of wiring patterns. (I looked up ones for autocomplete and they looked in the neighborhood. It takes all the Ins and Outs in any given rack and generates a weight table of probability.)
Then for whatever modules are on the rack, the jack placement would be weighted according to the likelihood a connection of A type jacks into a connection of B type as used in actual patches. Then you could have a knob where 0 is precisely the most likely, and 1 is what you have now (completely randomized), so you could get the level of randomness you want. If there's not enough data, then I think you'll want to generalize the grammar more; so it could make a model where roughly submods of type-A generally more often jack into submods of type-B, and add that to the weighting.