Skip to content
TopicTracker
From HackerNewsView original
TranslationTranslation

Free, on-device text-to-speech in the browser (and a scam-call sound board)

The article introduces free, on-device text-to-speech capabilities running directly in the browser, using Web Speech API or similar tools. It highlights practical use cases, including creating a scam-call sound board for pranks or awareness, while noting that all processing happens locally for privacy and speed.

Background

- TTS (text-to-speech) traditionally relied on cloud services (AWS Polly, Google Cloud TTS, ElevenLabs) that charge per character or per second and require an internet connection. This article explores fully on-device, free alternatives that run inside a web browser. - The key technologies covered: Web Speech API (built into browsers, voice quality varies by OS), and the newer Web Audio API + local neural TTS models (e.g., Coqui Studio, Piper, or other small transformer models that can run client-side via ONNX or WebAssembly). - "On-device" means no data leaves your computer — important for privacy, latency, and offline use. "Free" refers to both no monetary cost and no API-key dependency. - The practical use case demonstrated is a "scam-call sound board": loading multiple TTS voices locally to simulate a phone call, which shows real-time switching and low latency of the approach. - Mozilla's deprecated Web Speech API (still in browsers) versus newer WebGPU/WebNN approaches for running on-device AI models — this article sits in a transition period where browser-based local AI is rapidly maturing.

Related stories