Now that you have your voice, it’s time to generate some voiceovers! Head over to the Speech Synthesis tab and try out your voice - you can find them in the dropdown menu. You can preview all voices by pressing
► next to the voice name. If you don’t see the button, you need to make an initial generation for it to show up. Simply write some text in the text box and press generate. Amazing, isn’t it?
However, there are a few things to keep in mind:
- The settings of the AI are nondeterministic, so each time you press generate, you will get slightly different results.
- If you click the drop-down for Voice Selection and start typing, you will automatically search through your list of voices.
- The Voice Settings have a big influence on how the voice sounds and performs. Playing around with these settings will be necessary to find the best tone and performance for your project. A lot of users set them to fixed values that they know work for them, usually stability around 40 and similarity around 75, and then keep generating until they find a performance they like without changing the values much.
- The model you choose should be influenced by the language you want to use. If you are doing primarily English voiceovers, we strongly recommend sticking to the monolingual model since the multilingual model is still in its experimental stage and can have a few quirks that are being worked on.
- You can sometimes force the AI to deliver something in a specific way with a certain emotion depending on the text, context, and dialogue tag. For example,
"That's not funny", he said seriously.This is not perfect but it is a workaround that a lot of people have found helpful while we wait for the next iteration.