laughter-synthesis

Tailored laughter synthesis

Development

# Install development requirements
pip3 install -r requirements/dev.txt

# Automatically apply code format
make format

Virtual Environment

# Create venv
make init_venv

# Install dependencies
sudo apt install portaudio19-dev python3-pyaudio
pip3 install -r requirements.txt

Docker

# Build the development environment
docker-compose build

# Run the jupyter server
docker-compose up

Dashboard

Plotly dashboard is integrated as a playground for the audio synthesis. In the dashboard, the following options can be configured and new audio synthesized:

• Number of people in the crowd
• Female vs. male ratio of the crowd
• Individual intensity level of clapping, laughing, whistling and booing

# Change directory to the project root
cd audience-synthesis-playground

# Install requirements
pip install -r dashboard/requirements.txt

# Run the dashboard server
python -m dashboard.src.app

Links and Resources

Unsupervised acoustic unit discovery for speech synthesis using discrete latent-variable neural networks

• Github Author (Benjamin van Niekerk): https://github.com/bshall?tab=repositories
• Vector-Quantized Contrastive Predictive Coding: https://github.com/bshall/VectorQuantizedCPC
• VQ-VAE for Acoustic Unit Discovery and Voice Conversion:
  • https://github.com/bshall/ZeroSpeech
  • https://bshall.github.io/ZeroSpeech/