If your completed trained model is converting distorted, harsh outputs, review our training article here to make sure that your dataset is of sufficient size and quality.
In order to prevent distortion in your completed voice, there needs little no distortion in your dataset. To do this:
- If it has distortion in dataset, apply EQ and compression to reduce it
- Ensure the voice in the dataset has a range of volumes, but does not exceed -9 db
- Ensure your voice has enough low-mid range frequency and is not only high end frequencies