TTS Overlay Twitch Overlay

Viewers who never stand out suddenly get heard, just for typing one line. When a viewer types !say followed by some text, a plain card appears at the bottom with their name and avatar, and a real, natural-sounding voice reads the message aloud on stream. Speech runs fully local in the browser through a neural Piper voice model bundled right into the overlay: no cloud service, no server of your own, no Docker, and after the first load no further downloads either, even offline. A built-in pronunciation dictionary makes sure English names and streaming slang still come out right under German pronunciation rules, and channel-specific entries can be added on top. Requests are read one after another in a queue; paid StreamElements redemptions jump the free chat queue with a gold highlight, a single store item is all it takes to set up. Five card themes, from lime through coral and ice blue to a colourless mono look, cover most scenes, and a text-only light theme drops the card panel entirely, placing name and text straight over the video with a heavy drop shadow for legibility on any background. Long messages scroll smoothly upward instead of being cut off. Position, volume, speaking rate, character limit and who may use the command are all configurable. As an OBS browser source it is a single HTML file, with the voice model sitting in its own folder right next to it.