Text to Speech

This tool uses the Web Speech API built into your browser, which means the voices come from your operating system rather than from a server. Nothing you type is uploaded. The practical consequence is that the available voices differ by device: Windows supplies its installed SAPI voices, macOS and iOS supply the Apple voice set, and Android supplies Google's. A voice list that looks rich on one machine can be short on another, and a voice you choose is not guaranteed to exist for someone else opening the same page.

Some browsers load the voice list asynchronously, so it can appear empty for a moment on first visit and populate a beat later. If no voices show at all, the operating system has none installed; on Windows they are added under Time & Language settings, and on Android through the text-to-speech engine settings.

Most of the awkwardness in generated speech comes from the text, not the voice. Punctuation is what a synthesiser uses for timing: a comma produces a short pause, a full stop a longer one, and a paragraph break longer still. Text with no punctuation is read as one breathless run. Adding it back is the single biggest improvement available.

Rate is the next lever. The default is roughly natural conversational speed; slowing slightly helps with technical material and speeds above about 1.5x are useful for review but tiring to listen to. Beyond that, work around known weak spots. Abbreviations are unreliable, so write "Doctor" rather than "Dr" and "for example" rather than "e.g." if it matters. Numbers are read according to context the engine guesses at, so 1/2 may come out as "one slash two"; write "one half". Acronyms are sometimes spelled and sometimes pronounced, and spacing the letters (N A S A) forces spelling. Proper names, especially non-English ones, are frequently wrong and can often be fixed by respelling them phonetically.

Proofreading is the strongest use. Hearing your own writing read aloud exposes repeated words, missing connectives, and sentences that run too long, because the ear catches rhythm problems the eye skips over after the third re-read. Reading a draft back before sending it is a two-minute check that catches a surprising amount.

It is also genuinely useful for accessibility and for study. Students with dyslexia often comprehend substantially more when following along with audio, and hearing a language being learned alongside the written form reinforces pronunciation. Practical limits are worth knowing: browser speech is designed for reading aloud in the moment, not for producing a file, so there is no download here, and long documents are better broken into sections since some browsers cut off very long utterances. For published audio, a dedicated neural TTS service will sound markedly better than any operating-system voice.