[ruuk]

I hadn't noticed you linking to httpttsd, but that's cool, too. Once you're happy with your python server, I won't be at all offended if you want to just link to that. It's also been my observation that python stuff seems much easier to package up as standalone packages that people can install/run as proper windows services, and that'd definitely be a great goal (in time) for your python server.

I felt a little bad about basically replacing your perl script, but it seemed like a waste not leveraging all this backend code.
I do plan to eventually create a windows installer and make this work as a service. I've actually done this before, but that was about 8 or 9 years ago so I remember almost nothing

One other thought I literally just had was that it might be good to add an option into the config section that lets you toggle between screen-reading mode (what we currently have) and a non-screen-reading mode. The latter essentially being silent all the time, but leaving your addon available for other future addons to leverage it if they want their addons to generate speech, for example if someone wanted to write an addon to speak subtitles.

I've been thinking for a while that it would be nice to have this provide a module for other addons so they could just for instance import xbmcspeech and then xbmcspeech.say("some stuff"). This will probably work best If I split out the backends into the xbmc module and have the service separate, importing the backends.
I do like having the service be able to work all in one addon with no dependencies because it makes it easier to install directly from the zip, and I could actually have it provide a module without separating the backends, but then if you disable the addon, the module will no longer be available, and I want the service to be disabled when "off" both because it then no longer uses any resources and because that allows for it to be included in an XBMC install where it can disable itself on first run. This is the only way so far I have come up with that that will provide pre-installed speech without affecting sighted users.

---

It seems to be certain length that plays, rather than some fraction of the total length. Something like 1 to 1.5 seconds.
While I'm sure it's not directly related, it seems to happen to wavs of approximately the same length as the crashing issue.

Yes, I've had another look, and have found the cause of the truncation. There was a 256K limit in size (after channel mapping/resampling).
I have a fix for that on newclock3 and gotham_rbp_backports branches. It should appear in a future build.

Thanks for checking on that and making a fix. I really appreciate the time you've taken to follow this.