A small neural TTS running locally in your browser via Piper. The smallest voices are ~22 MB and synthesize faster than realtime on CPU. After first download a voice is cached in the browser’s Origin Private File System; subsequent uses are instant.

This is the companion to Ask a Tiny Mind. Same idea, different modality: small models, local-only, no backend.

Voices come from the Piper project (rhasspy). Inference runs in your browser via piper-tts-web + ONNX Runtime Web. Tip: Cmd/Ctrl + Enter in the textarea triggers synthesis.