Creating Your Own Singing Voice Synthesizer: Overcoming Data Collection Challenges

Explore the process of creating a personalized singing voice synthesizer in this 21-minute conference talk from the Audio Developer Conference. Delve into the challenges of data collection for custom voice datasets and learn various approaches to overcome them. Discover the advancements in deep-learning-based singing voice synthesis and the limitations of proprietary systems. Gain insights into open-source alternatives that allow for custom voice creation. Examine methods ranging from manual labeling to fully automated procedures for collecting phoneme-level timing and other crucial data points. Understand the current landscape of public datasets and their linguistic limitations. By the end of this talk, acquire the knowledge to embark on creating your own unique singing voice synthesizer, making this technology more accessible to audio enthusiasts and developers alike.