Post

I’ve had the webtospeech.app domain for years, maybe half a decade now. Today I’ve re-launched a service on it, which uses Gemini 3.8 Flash’s tts model. It pulls the main content from a webpage, and converts it into your own little audiobook.

Originally, I was using the neural tts services from Azure, AWS and Google, which were good enough for me, but weren’t worth offering as a service. When LLM’s appeared, long context TTS offering quality dropped through the floor, and the neural services stopped even getting a look in.