Option 1 is not feasible - instead of promoting one product (voice library) you’d have to promote 2 - library and the app.
Option 3 is not that bad if you have time. You can have an open source library with a free license + commercial license. If your lib is really good, it gets used in OSS projects and become known to commercial players at which point you can have a sale or two. But again - it will take much time.
Re: Option 2. I do not really believe you have to get a patent - I think the patent is important if you plan to sell to more than one vendor. However, IANAL.
What I would do? Hmm… I’d make a public demo page, comparing the same voice playback over a few of voice applications. You’d have to have some testbed network implementation that has configurable latency and packet drops and whatnot to record that demo. Then I’d send that demo to VPs of the existing voice applications.