If you have a large enough market, the simplicity and repeatability of dedicated controllers with buttons chosen precisely for your game's design and so on is attractive.
If you don't, you run into the problem that low volume production of such gear isn't going to make the price point any more attractive, and it's fairly bulky and expensive for something you can only play a few games with.
Anyone know what the feasibility might be of, instead, of taking advantage of what is already available? For mics, the attempt to make voice control a fad left a fair number of consoles already equipped with one, cellphones and tablets all have them and support wired or wireless headsets, and USB mics of unexceptional quality cover everyone else for not much money. On the guitar side, probably-awful 'beginner' units are $60-80(probably less if you get one used after buyer's remorse claims the original victim), and essentially any electric guitar will support putting out a low-level signal into a 1/4inch jack. If a device already has a line in, a simple mechanical adapter will do, if not, cables that are a USB audio-in on one end, 1/4inch jack on the other are quite cheap. Once you had that, your game could presumably crunch the guitar's output and (depending on how much 'game' and how much 'learning tool' you want) do anything from treating a few large contact areas as 'buttons' to actually grading you on the degree to which your results match the correct output.
I doubt that, if the user needs to purchase everything, particularly new, you could beat the package cost of a mass-produced controller pack; but if you don't think that you have the volume for a suitable production run of instrument-controllers, it seems like an approach that has very low marginal cost and can work with more or less any instrument floating around in the wild, might be less risky and more approachable.