Creating your own Raspberry Pi voice assistant is a fun way to create the smart home of your dreams. It also allows for a great degree of customization. You can choose how interactive the system is, integrate AI with tools like ChatGPT and even animate holographic figures to visualize your character. But if you've ever wanted to change how your voice assistant sounds, you should check out Erik Bjorgan's latest project, TextyMcSpeechy.
This open-source tool makes it possible to add a custom voice profile to your text-to-speech (TTS) project. According to Bjorgan, the project was created out of necessity. He was looking for a way to create a custom model of his own voice so that he could use it in a Raspberry Pi-based text-to-speech project. But when he looked around online, there wasn't an easy way to do this so he opted to develop a streamlined process that anyone can use.
TextyMcSpeechy will work for those who want to make their own TTS model from scratch or create one using samples of another person's voice. It works with Applio, an open-source AI system geared toward cloning voices for TTS uses. Bjorgan confirmed that he's using the setup himself to recreate celebrity voices in his various smart home devices so you can really go crazy with the use cases.
Bjorgan shared a complete list of features for TextyMcSpeechy over at the official project GitHub page. It includes things like the ability to create TTS models using your own vocal recordings or datasets put together by someone else. But it also mentions the ability to use checkpoint files to train TTS models faster while retaining good quality.
The GitHub page provides instructions on how to use TextMcSpeechy as well as details about what you need to create a TTS model. You will need a voice dataset of the subject whose voice you want to clone and then need to pair this audio with text so the system can build a profile of the voice. If your Raspberry Pi project needs a custom voice, check out TextyMcSpeechy and test it out for yourself.
