Noise in a voice model or low quality outputs in a trained voice is likely due to recording conditions of the dataset. provided. To ensure that there is no noise or low quality outputs, make sure that your dataset does not have any white noise or a constant hum.
Some more tips on reducing noise:
- Remove white noise/hum with a de-noise plugin in your DAW, like Izotope RX
- Make sure the room you’re recording in is soundproof, if you hear a slight reverb on your voice when recording, you need to provide sound proofing or get a microphone shield
- Ensure the recording equipment for your microphone is high quality