Invention:
This invention is a software system that helps generate a natural sounding synthetic voice from a small amount of audio files. The audio files can be from an individual’s own voice so that when their speech becomes later impaired, their same voice can still be generated.
Background:
Currently, there is technology available that is able to generate a natural sounding text-to-speech. However, these models require large sets of audio files and corresponding text (~24 hours). This poses an issue for ALS patients; ALS patients inevitably lose their ability to speak so it can be difficult to have this much data. By reducing the amount of data required for the model to 1.5 hours, it would relieve a huge burden from ALS patients.
Applications:
- Patients who need voice synthesizer
- Archiving parents and grandparents voices
- New voices for text to speech applications like audiobooks
Advantages:
- Less data required
- Natural sounding text-to-speech
