3 ms·
Hours of audio comprised of clean spoken sentences, zero noise, and uniform microphone quality is ideal. Some of the predominant base data sets used for transf
by echelon 5y ago
Hours of audio comprised of clean spoken sentences, zero noise, and uniform microphone quality is ideal.
Some of the predominant base data sets used for transfer learning, such as LJSpeech [1], are unfortunately noisy and non-uniform.
[1] https://keithito.com/LJ-Speech-Dataset/ https://keithito.com/LJ-Speech-Dataset/
- kragen 5y agoThank you very much!