Creating AI Voices

​

Setup your training data

Learn how to structure your training data to get the best results.

{
  "name": "Sample Model",
  "description": "A custom voice model ready for training",
  "audio_directory": "audio"
} 

​

Optimize your training data

Optimizing your training data will help you get the best results from your custom voice model. We recommend the following guidelines when preparing your training data files:

• Use dry acapella voice files - This means that the audio files should not contain any background music or sound effects (e.g. reverb, delay, etc.).
• Use high quality audio files - The audio files should be recorded in a quiet environment with minimal background noise (e.g. no traffic, no wind, etc.). Ideally the audio files should be recorded in a professional studio.
• Use a consistent volume level - The volume level of the audio files should be consistent across all files (e.g. no clipping, no distortion, etc.).
• Use a consistent sample rate - The sample rate of the audio files should be consistent across all files (e.g. 44.1kHz, 48kHz, etc.).
• Use a consistent bit depth - The bit depth of the audio files should be consistent across all files (e.g. 16-bit, 24-bit, etc.).
• Use a consistent file format - The file format of the audio files should be consistent across all files (e.g. WAV, AIFF, etc.).