Hal can use practically any SAPI 5 compliant TTS (Test-To-Speech) software. This is special software that has sampled or taken samples from human voices and converted them into a recognized digital format.
Diphone or half-phones would require a lot more work and the end result might still be plagued with pronunciation, inflection, timbre, and other errors. Their sampling methods as such do not always result in the best of quality whereas some TTS voices are almost indistinguishable from that of a human.
I hope this sort of answers your question. There are several companies that Hal could use aside from Microsoft's Mary or Anna like those from NeoSpeech, Cepstral, AT&T Natural Voices, etc.
Search through the Speech section of the forum here for a lot of additional info regarding TTS speech.
Lastly, Welcome to the forum!!