Creating custom voices
In the "Custom Voices" section of RadioNewsAI, you have the ability to create a clone of your own voice or that of an existing newsreader where you have permission and the rights to do so.
Custom voices can be created from a one-minute audio sample, but you can upload up to five minutes of audio to give the system more source material.
Keep in mind that these custom voices are instant clones. They are not perfect duplicates, and results can vary. Most of the time, the outcome is at least a credible voice trained as a newsreader, even if it doesn't sound precisely like the original.
Here are some key considerations when using samples to clone a voice:
- Ensure using a high studio-quality recording, as the AI replicates the quality found in the samples.
- Avoid including any background music or sound.
- Ensure the samples feature actual news reading, as the AI will adopt this style.
- Minimize breathing sounds or utilize a noise gate to diminish some breathing noises.
- Avoid leaving white spaces between news stories. Read the entire newscast as one or edit out the white spaces to prevent the AI from inserting unexpected white spaces in the output.
Accents can sometimes be challenging to clone and, for unspecified reasons, may depend on the individual. Therefore, if the accent for a voice isn't what you’re seeking, consider trying another voice.