About

This project actually started off as a joke. Me and my friends like to use a TTS bot so that if someone can’t speak, they can still take part in conversation. I find that it’s easy for people to ignore mute users because you have to actively monitor the chat. When using a TTS bot, people can better participate in conversation because you can actually be heard.

The only problem is that it becomes more difficult to figure out who is speaking, especially when multiple people are using TTS. Hearing a distinct and recognizable voice allows you to subconsciously link the voice to a person at any time in the sentence.

All of the actual AI processing is from a project called coqui. It’s a super capable toolkit and all credit for the voice processing goes to them. To actually make it clone a voice, it needs a reference recording. Any recording will do, and you only really need between five and twenty seconds of audio. The bot automatically uses the correct reference based on the command sender. That way, only you can use your voice.