We can now put unseen sentences through the system, to perform synthesis.
Edit your config file to turn synthesis on
NORMLAB : False MAKECMP : False NORMCMP : False TRAINDNN : False DNNGEN : True GENWAV : True CALMCD : False
and run this step
$ python /Volumes/Network/courses/ss/dnn/dnn_tts/run_dnn.py feed_forward_dnn_WORLD.conf
You will now have synthesised waveforms in gen
for your validation and test files.
At this point, we have a system that can only synthesise held-out sentences. This is because we are using the durations from forced alignment. To make this a fully-functional TTS system, we will need to add a duration model (but that is left for another day).