Extract navigation state out of SSML <marker>s.
Output audio cues in live speech.
Turned out I was overcomplicating things, all I needed to do was use file paths not URIs.
eSpeak NG handled it from there.
Allow disabling of URL rewriting for SSML output.
Use libsndfile to write audio files.
Import libsndfile, prepare to use it for writing audio files.
Output converted webpage.
Partially fix segfaults, enough to continue on.
Still questions remain, and there's broken C APIs exposed.
Switch to C for I/O. FIXME currently segfaults.