Consider a separating the goals.
OTR is much more successful because it allows ad-hoc connections instead of trying to make authenticity guarantees. It encourages authentication, but the decision of when to invest that extra effort is left with the user. Fundamentally, OTR just offers privacy from eavesdroppers, and it does that well. Every other feature is extra. Even better, these new features (e.g. authentication) don't have to start from nothing. A big benefit of separating features is that is added later can leverage the existing infrastructure.
I personally believe the "killer crypto app" that will actually be adopted is a tool that is the opposite of OTR. It should only promise ad-hoc authenticates and do it in an generic style. I mean something that is somewhere between "gpg --clearsign" and 2channel/4chan style tripcodes[1]. In an ideal world, this needs to have a clearsign-style mode, so it can be used on existing forums. If the browser simply offers a "sign this" button next to textareas (no password, auto-gen they key and stuff it on a keyserver, just make it happen), it will be used. On the decode side, any blocks of text the browser sees that look like a signature get check, with the key pinned. That's it. Any features that allow nicer integration with email, forum software, etc, are nice but optional.
This provides a minimum, anonymous level authentication (forgery prevention), and the only UX the user should have to see is the "sign this" button and a loud "possible forgery" warning if a signature doesn't match the pinned key. It even works regardless of who is transporting the message (different websites, email, gopher, avian carriers) without needing an "identify provider". Upgrading this infrastructure of pinned keys into something more interesting can happen later on, if desired.
Would this work? I believe it's worth trying. It is clear that trying to solve all of the security needs in a single tool has not worked, and the one tool that got any adoption at all (OTR) has a much more narrow set of goals.
[1] http://en.wikipedia.org/wiki/Imageboard#Tripcodes