Text-to-speech extensions reviewed: Speechify, NaturalReader & Read Aloud

Here's the thing nobody says out loud about read-aloud extensions. Before one can speak a page to you, it has to read that page. All of it. And if the voice coming out of your laptop sounds like a radio presenter rather than a 2009 satnav, that voice almost certainly wasn't generated on your laptop.

Which means the text went somewhere.

That's the whole review, really. But the three big text-to-speech extensions answer the question in three different ways. I lined up Speechify, NaturalReader and Read Aloud on two things: how much of your browsing can this thing see, and where does the text physically become audio?

Quick verdict

• Speechify — cloud by design. Its own privacy policy says content you save is stored on your device and synced to Speechify's servers, processed and stored in the US. Great voices, wrong shape for a confidential contract you're too tired to read.
• NaturalReader — also a cloud service, but with the most explicit compliance paperwork of the three: GDPR alongside FERPA, COPPA and New York Ed. Law 2-D. That matters if you're deploying to students.
• Read Aloud — open source and free. It uses the voices already on your machine, and only reaches a cloud if you plug in your own API key. The privacy-conscious pick, and the least polished-sounding one.
• The built-in option beats all three for sensitive documents. Chrome and Edge both read pages aloud with nothing installed.
• My take: pick by document, not by voice. Public article → any of them. Anything you'd hesitate to paste into a random web form → local voices only.

What text-to-speech extensions actually do

Two separate jobs, and people conflate them constantly.

Job one: get the text. The extension pulls the article body out of the page, skipping nav bars, cookie banners and ads. That means running code inside the page you're looking at — the scripting capability and host access doing their work. The mechanism is identical whether the extension is honest or not.

Job two: turn that text into audio. Here the category splits, and it's invisible from the UI. Some extensions hand the text to a speech engine already installed on your operating system. Others POST it to a server, get an audio stream back, and play it.

Same button. Completely different privacy story.

The tell: local voice or network voice

The web platform encoded this distinction years ago. Every voice exposed through the Web Speech API carries a localService flag — true when a local synthesizer makes the audio, false when a remote one does. MDN spells out why the flag exists: remote voices can carry extra latency, bandwidth and cost.

Read that as a privacy signal, because that's what it is. Latency and bandwidth are what "your sentence made a round trip to a data center" feels like from the outside.

So here's my rule. If a voice needs the network, your text is on someone's server. If it works in airplane mode, it isn't. Two-second test, more reliable than any marketing page.

Security analysis: how much can it see?

Page access is the other half. A read-aloud extension can be built two ways. It sees the current tab only when you click it — or it runs a content script on every page whether you asked for audio or not. Both ship in this category, and the install prompt is where you find out which one you got.

Chrome's permissions reference decodes the strings. I've also written up what "read and change all your data on all websites" actually buys an extension. That warning is the most important line in this whole category.

Whatever you install, you can tighten it afterwards. In chrome://extensions, open Details, find Site access, and switch it to On click. The extension then sees nothing until you press its button.

For a tool you use a few times a day, that costs one click and removes a standing window into your banking tab. It also answers an awkward question: what happens when a well-behaved extension changes hands or ships a bad update? That's a pattern I keep hitting with utilities people installed years ago and never looked at again.

Speechify

The best-known name here, and the most ambitious about what it will read. Its Chrome Web Store listing pitches Google Docs, PDFs, webpages and books, in a large voice catalog, at up to 4.5× speed. It claims more than 50 million users — a marketing figure rather than an install count, but the product is popular and the voices are genuinely good.

Those voices are the giveaway. Generating them takes real compute. Speechify's privacy policy describes the architecture plainly: content you save is stored locally and synced with the company's servers, and information submitted to Speechify is transferred to, processed and stored in the United States.

That's disclosed, not hidden, and it's the only way to ship voices of that quality. But combine the feature list with the architecture. "Reads your Google Docs and PDFs" plus "syncs to our servers" means the quarterly figures and the draft contract get read by something other than your browser.

To be precise about what's sourced: the policy covers saved content and US processing. The synthesis path itself isn't spelled out — I'm inferring it from that plus the voice quality. Either way, if your answer to "would I paste this into a stranger's web form?" is no, this isn't the tool for that document.

One more thing. Speechify is account-based, so a persistent identity is attached to the text you send. That's the difference between a service that processes data and one that accumulates it. The privacy tab on the store listing is the shortest route to the developer's own declarations.

NaturalReader

Same fundamental model — a cloud service with a browser front end — but a different posture around it. NaturalReader's help center privacy article states that registration needs only an email address, that the company doesn't sell user information to other marketers, and that non-personal data such as browser and OS details is collected when you use the site.

The part I'd actually weigh is the paperwork. NaturalReader documents compliance with FERPA, COPPA and New York Ed. Law 2-D alongside GDPR, and keeps a separate privacy and data security collection for its commercial version.

That reads like a vendor that's been through school-district procurement. It says nothing about whether the text leaves your machine — it does — but it says something about who's accountable afterwards. Check the store listing for the current permission set before installing.

Read Aloud

The outlier, and my pick for anything sensitive. Read Aloud is open source on GitHub, maintained by developer ken107. Its default mode uses the voices your browser and operating system already provide: synthesis on your machine, no account, no upload, no fee.

Cloud voices are available, but the design is inverted. The premium voices documentation covers hooking up Google WaveNet, Amazon Polly, IBM Watson and Microsoft voices with your own provider credentials. If text goes to a cloud there, it goes to your cloud account, on your terms, because you chose it.

The developer states on the project's site that the extension asks only for the minimum permissions it needs. I'd still check rather than take anyone's word for it. The Firefox add-on listing and the Chrome listing both show requested permissions before you commit.

The honest downside: OS voices sound like OS voices. If you're getting through 200 pages a week, the polish gap is real.

Where they land

	Where audio is generated	Account required	Source available	Best fit
Speechify	Company servers, US-processed per its policy	Yes	No	Long public reading, best voice quality
NaturalReader	Company servers, documented education/GDPR compliance	Yes	No	Schools and organizations that need a paper trail
Read Aloud	Your device by default; your own cloud keys if you opt in	No	Yes	Sensitive documents, privacy-first setups

Two axes, and they don't rank the same way. Blast radius is about page access, and it's a per-build question: check site access, set on-click, recheck after updates. Content exposure is architectural and doesn't change. A cloud voice means the text left, every time, by design.

That's the same split I found in the PDF extension category, where the extension with the widest permissions wasn't the one that shipped your file furthest. Want both dimensions for a specific listing? The text-to-speech category in the Extenshi catalog shows declared permissions and version history together.

Alternatives worth knowing

• Chrome's own read-aloud. Reading mode on desktop has a play button; Google documents how to switch it on and keeps a broader page on hearing text read aloud. Voices come from your OS, so quality is on you — but nothing gets installed and nobody gets page access.
• Edge's Read Aloud. Built in, handles PDFs. Microsoft's feature page notes it works online and offline, with only a subset of voices available offline. A voice that needs a connection is rendering your text elsewhere, so for a confidential document, pick from the offline set.
• Better OS voices. Windows and macOS both ship extra high-quality system voices you can install from settings. They run locally, and they close much of the quality gap that pushes people toward cloud extensions.

Common questions

Is there a Chrome extension for text to speech? Several — and Chrome has one built in, so reading mode's play button needs no extension at all. Try that first.

Is text-to-speech free? The browser built-ins are, and Read Aloud is free and open source. The cloud services run freemium and their tiers move around, so check pricing on the vendor's own site rather than a listing screenshot.

What's the best text-to-speech extension for Chrome? It depends on the document, which is the least satisfying answer and the only correct one. Best voices: Speechify. Best paper trail: NaturalReader. Best for anything confidential: Read Aloud on local voices, or your browser's built-in reader.

What about Firefox, Edge, Safari or Opera? Read Aloud ships a Firefox build on AMO, and Edge has its own reader built in. Availability elsewhere changes, so check the add-on store for your browser — that listing is the only one telling you what that build requests.

Is Speechify safe? Safe from what matters here: it's a real product with a published policy, not malware. But that policy describes syncing saved content to the company's servers, so "safe" and "private" aren't the same question. Only the second one depends on which document you feed it.

Final recommendation

Match the tool to the document, not to the demo.

For public reading — news, blog posts, documentation, anything you'd happily forward — Speechify's voices are the best experience here, and the privacy cost is text you were reading in public anyway. NaturalReader is the one I'd hand to an institution, because its compliance documentation is written for people who have to answer for it.

For anything confidential — client work, legal documents, medical letters — use local synthesis. Read Aloud with browser voices, or the reader your browser already ships. The text never leaves, and the question stops being interesting.

Then do the two-minute hygiene pass on whichever one you keep. Set Site access to On click, and recheck the permissions after a major update. This category is full of tools people installed for one semester and left running with full page access for years. If you've never counted yours, browse what's actually installed.

See the security report for your text-to-speech extensions →

Sources

• SpeechSynthesisVoice: localService property — MDN
• Web Speech API — MDN
• Chrome extension permissions reference — Chrome for Developers
• Speechify Privacy Policy
• Speechify — Text to Speech — Chrome Web Store
• Speechify — privacy practices — Chrome Web Store
• Privacy — NaturalReader Help Center
• Privacy & Data Security, commercial version — NaturalReader Help Center
• NaturalReader — AI Text to Speech — Chrome Web Store
• Read Aloud — source repository on GitHub
• Read Aloud premium voices documentation — GitHub
• Read Aloud project site
• Read Aloud — Firefox Add-ons
• Read Aloud — Chrome Web Store
• Use reading mode in Chrome — Google Chrome Help
• Hear text read aloud — Google Accessibility Help
• Read aloud — Microsoft Edge features

---

This article is based on publicly available documentation published by the vendors named, plus official browser and web-platform references. Extenshi does not independently verify all vendor claims. References to specific companies or products reflect those companies' own published policies and do not constitute accusations of intentional wrongdoing. If any entity referenced in this article believes information is inaccurate, contact [email protected] and we will review and update.