This uses the voices already on your device
The tool is a front end for the Web Speech API, the speech synthesis feature built into the browser. When you press play, the text is handed to your operating system or browser speech engine, which produces the audio and plays it through your speakers. No text is sent to a cloud service, no neural voice model is downloaded, and no account or key is involved.
The direct consequence is that the voice list is not ours. It is whatever your system exposes, which is why the same page offers dozens of voices on one machine and two on another. Each entry shows a voice name and a language tag, and quality ranges from the highly natural voices shipped by recent operating systems to older formant synthesizers that sound plainly robotic.
Some browsers populate that list asynchronously after the page loads, so the dropdown can start empty and fill in a moment later. If it stays empty, your platform is not exposing any voices to the browser, and installing a system voice pack is the fix rather than anything in this page.
Rate and pitch
Rate runs from 0.5 to 2.0 where 1.0 is the natural speed of the selected voice, so 0.5 is half speed and 2.0 is double. Slower helps with unfamiliar terminology and with dictation-style listening; faster is useful for proofreading a draft you already know.
Pitch runs from 0 to 2 where 1.0 is the voice default. Moving it changes the perceived fundamental frequency, and the effect varies a great deal between engines. High quality neural voices often respond to pitch changes poorly or ignore them, because the voice was trained on natural speech rather than synthesized from parameters. Older synthesizers respond strongly and can be pushed to sound unnatural.
Both controls apply when playback starts, so changing a slider mid utterance does not affect the sentence that is already being spoken. Stop and play again to hear the new setting.
Reading your own text back to you
The most useful application of a tool like this is proofreading. Reading your own writing silently, your eye supplies words that are not on the page and glides over repetitions. Hearing it read aloud by a voice that follows the punctuation exactly exposes run-on sentences, a missing word, an accidental duplicate, and clauses that turn out to be unreadable when spoken.
It is also a quick accessibility check on copy you are about to publish. If a sentence is hard to follow when spoken at a steady pace, it is hard to follow for anyone using a screen reader.
Punctuation is doing real work here, because the engine derives its pauses and intonation from it. A paragraph with no commas is read as one long breathless run. That is a fair signal about the writing rather than a limitation of the tool.
A worked example and the download question
Paste three paragraphs of a draft, roughly 400 words. Pick a voice matching the language of the text, set the rate to 0.9 so you can follow along against the screen, and press play. The whole passage is read as one utterance. Stop cancels it immediately; there is no pause and resume, no seek, no progress indicator and no word highlighting to follow along with.
The important limitation is that there is no way to save the audio. The speech engine renders directly to the audio output, and this page does not capture that stream, so there is no MP3 or WAV download and no way to use the result in a video, a podcast or a presentation. For a file you can keep, you need a text to speech service that returns audio, or a screen recording that captures system audio on your own machine.
Voice availability also differs per language. Selecting a voice whose language tag does not match your text produces the text read with the phonetics of the wrong language, which is occasionally comic and rarely useful.
Browser support and privacy
If the browser does not implement speech synthesis, the page says so rather than offering controls that do nothing. Support is broad across current desktop and mobile browsers, but the behaviour differs in the details: how voices are enumerated, whether a voice is generated locally or by a platform service, how long an utterance may be, and whether a backgrounded tab keeps speaking.
That last point is worth stating carefully for privacy. Speech synthesis on most desktop platforms is performed locally, but some browsers route certain high quality voices through a platform service on the vendor infrastructure. Whether your text stays on the device therefore depends on the voice you pick and the platform you are on, and this page has no way to tell you which case applies. Do not paste confidential text into any speech synthesis field, here or elsewhere, if that distinction matters to you.
Nothing you type is sent to this site or stored anywhere by this page. The text lives in the browser tab for as long as the page is open and is gone on reload.