Offline Text to Speech: 7 Free Tools That Work Offline

Why offline TTS suddenly got good

For years, offline voices sounded like a robot reading a phone book, and the good voices all lived in the cloud. Two things changed. First, small neural models got good enough to run on an ordinary phone CPU — Phonon is about 65 MB and starts speaking in 28 milliseconds, according to Gradium's own benchmark chart on the Hugging Face Open TTS Leaderboard. Second, those models are now open: Phonon's small English model is Apache 2.0, which means free even for commercial use.

As Gradium put it in its launch post on Hugging Face: "It needs no server and no network, and what you type stays with you." That last part is the real reason many people want offline TTS — not the plane, but the privacy. A script, a medical note, or a client document never leaves your machine.

1. Gradium Phonon — the new quality leader

Phonon is a small open-weight model (about 65 MB for the English version, 75 MB for the multilingual one) that runs on phones, laptops, single-board computers, and even in a browser tab. It speaks English, French, German, Spanish, and Portuguese, ships with more than 100 built-in voices, and can clone a voice from a short recording or design one from a text description.

The numbers behind the hype are real and checkable: Phonon was downloaded more than 200,000 times in its first two weeks, according to Gradium's follow-up post on Hugging Face on November 5, 2026. On the same leaderboard, it ranked first for intelligibility in German and Spanish and in the top three in every language it was tested in, next to models up to 100 times its size. Its 95th-percentile time to first audio on CPU was 28 milliseconds — Pocket TTS, the next open model, took 165 milliseconds.

The catch: Phonon is a model, not a finished reading app. Developers get a Rust runtime and a Transformers integration; everyday users will meet it inside apps built on it, not as a big friendly "paste text here" button. If you just want to listen to a document today, options 3 to 7 below are faster.

2. Piper TTS — the proven open-source workhorse

Piper has been the default answer to "free offline TTS" for a couple of years, and it still earns the spot. It is a small neural engine that runs entirely on CPU — even on a Raspberry Pi — with voices covering 44 language and region combinations. Install it with one command, download a voice, and it turns text into a WAV file in a fraction of a second; one recent test generated a 6.7-second sentence in 0.16 seconds.

The catch: it is a developer tool. You run it from the command line or a small local web server, and the voices, while clear, sound flatter than Phonon's. But it is free software under GPL-3.0, there is no account and no usage meter, and your text never leaves the machine. For bulk work — turning a whole book into audio, or narrating files on a server — it is still the most practical free choice.

3. OfflineTTS — offline voices in a browser tab

OfflineTTS runs AI models directly in your browser using WebGPU or WebAssembly, with five TTS engines — Kokoro, Kitten, Piper, Supertonic, and Pocket TTS — plus Whisper for speech to text. Once the model file is cached (the English Kokoro model is about 305 MB), synthesis happens on your device, with 117 voice options across 32 languages.

The catch: "offline" starts after a large first download, and non-English Kokoro voices still use an online service for one step before the audio is made locally. So it is offline in practice for English, and mostly offline for everything else. Still, for someone who wants to try local voices without installing anything, it is the easiest front door.

4. Readalix — the Chrome extension that reads pages offline

Readalix is a Chrome extension that reads web pages, PDFs, Docs, and Word files aloud while highlighting each word. Its browser voices are free with no limit, and more than 100 natural voices in over 30 languages run on your device after a small one-time download — so a read keeps going with no internet connection.

The catch: the free allowance for its online neural voices is 20,000 characters a day (roughly 14 pages), and removing that limit costs $15 a year or $40 once. The offline browser voices have no such limit, so the free path is real — it just sounds less polished than the paid voices.

5. Your phone's built-in voices — free, already installed

This is the option most people overlook because it is already on their phone. On Android, Select to Speak reads any text on screen once you turn it on under Settings > Accessibility, and downloaded voices keep working offline. On iPhone, Spoken Content under Accessibility > Spoken Content does the same, and the vTTS app adds offline EPUB and PDF reading in 17 languages with audio export.

The catch: system voices are built for accessibility, not performance. They read well, but you cannot export an MP3 from them, and the voice choice is whatever your phone shipped with. For listening to articles on the go, though, nothing beats free and already installed.

6. ToBe SAID — an offline voice engine for Android

ToBe SAID is different from the apps above: it is a voice provider, not a reader. Install it, set it as your preferred text-to-speech engine in Android settings, and every app that uses the system voice — TalkBack, ebook readers, accessibility apps — speaks with its natural AI voices, entirely on your device. It ships four natural English voices and two voices in each of 14 other languages, with no ads and no in-app purchases.

The catch: it replaces your system voice rather than giving you a place to paste text. If you want to turn a specific document into audio, you still need a reader app on top of it. But for anyone who reads everything through their phone, it upgrades every app at once.

7. Loudify — a simple offline reader for Android

Loudify is a free reader app that opens PDFs, ebooks, and documents and reads them aloud with no account, no ads, and no data collection. Unlike most reader apps, it is fully offline and private — text never leaves the device, with no subscriptions, no word limits, and no sign-up wall.

The catch: it leans on your phone's system TTS engines for its voices, so the sound quality depends on what voices your phone has installed. Pair it with an engine like ToBe SAID and it gets noticeably better.

The seven tools at a glance

The honest line: when offline is the wrong tool

Offline tools win on privacy, cost, and no-signal reliability. They lose on convenience. You install something, download a voice, and accept a slightly flatter voice than the best cloud models. And none of the free offline options hand you a polished MP3 in one click without setup.

So pick by the job. If the text is private, the internet is absent, or the volume is huge, go offline — start with Phonon if you build things, Piper if you process files, or your phone's own voices if you just want to listen. If you need a downloadable MP3 right now with no install — a voiceover for a video, narration for a presentation, audio for a course — a browser tool is faster. Our free text to speech tool does exactly that: paste your script, pick a voice, and download the MP3 in your browser, no sign-up and no credit card. And if your starting point is a document rather than typed text, our guide to apps that read PDFs aloud covers the PDF-to-audio route step by step.