Computer text-to-speech (TTS) is a significant part of Automaton, and today I began the process of adding such voices to the track.
Now and then I delve into the Mac TTS engine for material; an example is the Dorian Zoyd track 'Nine eyes'. For Automaton the TTS voice is used to provide some context to the process our protagonist, Dorian himself, is undergoing. As a sonic material it also allows my vocal part to be broken up, providing more diversity to the song.
The workflow for integrating the TTS voice started with selecting the most appropriate voice from the Mac's TTS preference pane (you may need to download extra ones if the defaults aren't to your liking) and setting up the text to synthesize in Textedit – usually the text needs to be made more phonetic to get around pronunciation hiccups (and don't forget that punctuation counts too!).
Next, I used Rogue Amoeba's Audio Hijack to capture the system's audio output as a WAV file. I then messed with the timings of the spoken parts in TwistedWave, adding silences to separate out individual sections and removing gaps that were too long.
I decided this time around to load that single speech file into a software sampler and set each zone up with a different speech segment, which Logic's EXS24 was woefully inadequate for (largely because this ageing sampler doesn't allow the same file to be loaded into different zones). Fortunately, it was Maschine to the rescue! I remembered that Maschine can be loaded up as a plugin, so tried that out for the first time.
With 12 speech segments loaded and zoned in Maschine (possibly enough for the first section of computer voices) I called it a day. Join me in a future CreativePact post where I (shock!) sequence and (horror!) mix these parts into the main tune.