Most of the time when you train a voice and it doesn't come out like you were expecting, there is either an issue with the dataset you used to train the voice or with your input audio file when trying to use the voice. Follow the below troubleshooting guide.
- Verify that your voice training dataset meets our recommended training quality standards. Examples of this include: totally dry, monophonic audio; at least 10-15 minutes, but the more the better; no instrumentals, background noises, or anything but pure voice data; no effects, like reverb or harmony; and more. Read our article here to make sure you meet all of the training criteria.
- Verify that your input audio meets our recommended input audio conversion standards. Like with voice training data, your input audio needs to be totally dry, clean, monophonic audio. Read our article here to make sure you meet all of the conversion criteria.
- If your training audio and input audio meet all criteria listed in the docs, and you are still unsatisfied with your audio quality, fill out a bug report in our help center - attaching the files you used for training and for conversion - and we will personally walk you through improving your results.