FAQ

Quick answers to the common questions.

Is it really free, with no word limit?

Yes — Quick TTS is free with no character cap, no sign-up, and no paid tier.

• Paste a tweet or paste a novel; it's the same tool.
• Display ads cover the hosting cost.
• There's no premium voice locked behind a paywall.

What file formats can I read aloud?

Quick TTS reads eight document formats locally in your browser, with no upload to a server.

• PDF, DOCX, and EPUB — covers the bulk of real documents and ebooks.
• ODT, RTF, HTML, TXT, and Markdown — handles everything else.
• Files are parsed entirely in your browser using lazy-loaded libraries (pdf.js for PDF, mammoth for DOCX, fflate for the zip-based EPUB and ODT formats).

For the deeper write-ups: how PDF parsing handles untagged reading order and what to do about scanned PDFs, and how to source DRM-free EPUBs and what to do about Kindle/Kobo files.

Do you store the text I paste?

No — your text never reaches our servers because all synthesis happens in your browser.

• Quick TTS is fully client-side; we have no logs of your text because it never leaves your device.
• Your most-recent text is saved in your browser's localStorage so it survives an accidental tab close.
• Clearing your site data wipes that local copy.

What's the difference between the default voice and "AI Voice"?

Browser TTS uses your operating system's built-in voice; AI Voice is three opt-in neural engines that run locally on your device.

• Browser TTS: instant, works everywhere, sounds like a standard system voice.
• Piper: ~60MB WebAssembly engine, runs in any modern desktop or mobile browser.
• Kokoro HQ: ~80MB, runs on your GPU via WebGPU (desktop Chrome/Edge). English only.
• Supertonic HD: ~380MB, 44.1kHz across 31 languages — the quality ceiling. WebGPU only, with no WebAssembly fallback.

For the engineer-y comparison — model sizes, latency, real chunk values, and the fallback chain — see Web Speech API vs Piper vs Kokoro. For an honest ranking of the actual voices each engine ships, see Best Free TTS Voices in 2026.

Does the AI voice send my text to a server?

No — Piper, Kokoro and Supertonic all run entirely in your browser.

• Piper synthesizes via WebAssembly; Kokoro and Supertonic synthesize via WebGPU.
• The only network requests are the one-time library and model downloads from jsDelivr and Hugging Face.
• After the first load, the AI engines work offline.

Does it work on mobile?

Yes — mobile gets Browser TTS, which works on every phone using its built-in voices, and Piper now runs on mobile too.

• Browser TTS handles the basic mobile use cases without any download.
• Piper-WASM works on current Android and iOS browsers (the prior iOS issues were resolved by Safari/Chrome updates).
• Kokoro HQ and Supertonic HD remain desktop-only because they need WebGPU, which mobile GPUs don't run reliably yet.

Can I use the audio commercially (YouTube, podcasts, products)?

Yes — commercial use is permitted on every engine, but the terms differ: some Piper voices require attribution, and Supertonic HD carries use-based restrictions the others don't.

• Browser TTS: governed by Microsoft, Apple, or Google's system-voice terms; check their docs at scale.
• Piper voices: LibriTTS-R and VCTK are CC-BY 4.0 (attribution required); Joe is CC0 / public domain.
• Kokoro HQ: Apache 2.0 — commercial use permitted, attribution recommended.
• Supertonic HD: different from the others — the model weights are BigScience OpenRAIL-M, not a permissive licence. Commercial use is allowed and no rights are claimed over the audio you generate, but OpenRAIL-M adds use-based restrictions that you must follow and pass on to anyone you share the model or its output with. The inference code is MIT.

We don't claim any rights over the audio you generate. You're responsible for complying with the underlying model licenses in your specific use case.

How long can my text be?

As long as you want — there's no character limit.

• The tool chunks input and streams playback, so a 50,000-word document plays the same as a sentence.
• Audio starts within seconds; the rest generates while you listen.
• AI Voice pipelines generation with playback so there's no long wait up front.

Why does the first AI Voice play take so long to start?

The first play of an AI engine downloads the model file once — every play after that is instant.

• Piper is ~60MB; Kokoro HQ is ~80MB.
• Download takes a few seconds to a minute depending on connection.
• After the first load, the model is cached in your browser.

I'm on a work computer and AI Voice is stuck at "Downloading..."

Your corporate network is almost certainly blocking jsDelivr or Hugging Face — the CDNs that host the AI models.

• If neither AI engine finishes initializing, that's the cause about 99% of the time.
• Browser TTS doesn't make external requests, so it always works on locked-down networks.
• Ask IT to whitelist cdn.jsdelivr.net and huggingface.co if you need AI Voice.

Why do I see ads?

Ads are how Quick TTS pays for hosting, the domain, and ongoing maintenance — that's the trade for keeping the tool free with no limits.

• No paid tier exists, so ads are the only revenue.
• An adblocker is a legitimate solution; we'd rather you use the tool than not.
• Ad slots are fixed display banners — no pop-unders, interstitials, or fake download buttons.

Is my keystroke data tracked?

No — Quick TTS doesn't run keystroke logging or any input-level analytics.

• Page-level analytics (Google Analytics 4) records visits, not keystrokes.
• Do-Not-Track is honored, so analytics is disabled if your browser sends that header.
• Full detail is in the Privacy Policy.

Something broke / I have feedback

Email hello@quick-tts.com. Include your browser and OS.