Download Speect and look at Building and Installing Speect or get the Debian packages (install the libspct1.0, libspct-py-native1.0, libspct-plugins-base1.0 and python-speect packages). Download a voice from http://hlt.mirror.ac.za/TTS/Speect/.
Unpack the voice in some directory:
tar -xzf cmu_arctic_slt-1.0.tar.gz cd cmu_arctic_slt
This voice does not have extensive NLP modules. Therefore, text to be synthesized must be plain text (normalized). For example, “we want one goal” and not “we want 1 goal”.
Run the Python interpreter:
Import the Speect module:
>>> import speect
Import the Speect audio module for playing utterances:
>>> import speect.audio
Load the voice config file:
>>> voice = speect.SVoice("voice.json")
Synthesize some text:
>>> utt = voice.synth("this is speect from within python")
Play the synthesized audio:
See speect.SUtterance.play() for more information if playback fails, also if speect.SUtterance.play() uses pyaudio then some errors might be printed. The WAV can still be saved to file even if errors occurred, see below.
View the utterance structure:
>>> print utt
Save the utterance by first importing the audio riff module:
>>> import speect.audio_riff
Get the utterance’s audio feature:
>>> audio = utt.features["audio"]
and save it to a WAV format file: