https://github.com/harrytwigg/feeg3003

Using DDSP to learn and synthesize singing with modifiable latent space parameters
https://github.com/harrytwigg/feeg3003

ddsp machine-learning ml

Last synced: 3 months ago
JSON representation

Using DDSP to learn and synthesize singing with modifiable latent space parameters

• Host: GitHub
• URL: https://github.com/harrytwigg/feeg3003
• Owner: harrytwigg
• Created: 2022-04-08T15:28:54.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2022-06-09T10:26:00.000Z (over 3 years ago)
• Last Synced: 2025-02-09T23:26:58.732Z (8 months ago)
• Topics: ddsp, machine-learning, ml
• Language: Jupyter Notebook
• Homepage:
• Size: 65.8 MB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Using DDSP to learn and synthesize singing with adjustable latent space parameters

Thesis project by Harry Twigg at the Univeristy of Southampton.

## Getting Started

Other reports or can be seen in their respective folders.

Results can be found in the folder `results`.

## Project Code

All experimental research can be seen in the folder `vocal_ddsp`. This contains pre-processing, training and evaluation scripts.

Dataset preparation, training and inferencing was done in Google Colab, the notebooks of which are available in the `vocal_ddsp/notebooks` folder.

Required gin configs are available in the `vocal_ddsp/gin` folder.

Training models and datasets:

https://drive.google.com/drive/folders/1CWdEco9tJ5fNBZfvLGqbSrfdYQ6pLZsz?usp=sharing