Diphone synthesizer

Basic concatenative speech synthesizer in English.

Overview

Basic speech synthesizer to convert input text to a sound waveform containing intelligible speech. It uses a very simple unit selection and waveform concatenation system, with the acoustic units being individual recordings of diphones.

• sympleaudio.py contains an Audio class that acts as an interface with the audio hardware, enabling operations such as saving, loading and playing .wav files.
• synth.py is the main program and contains the following classes:
  • Utterance: Performs basic word tokenization and normalization of the input text, converts words to phonemic transcriptions and extracts a sequence of diphones.
  • Synth: Takes in a sequence of diphones, reads the contents of their corresponding .wav files into Audio objects, and concatenates them, allowing to play and/or save the output as another .wav file.
• directory: path to the directory that contains the .wav files for diphone sounds - diphones go from the middle of one speech sound to the middle of the next one, capturing the transition between both sounds.

Instructions

Required arguments

• --text: text to be synthesized.

Optional arguments

• --diphones: directory containing .wav files (./directory by default).
• --play (default): play synthesized output.
• --save: output .wav filename.

Example

python synth.py --text "I sound like a robot" --save robot.wav --play