Narrator of Mordor

Reads web pages aloud with private voices that run on your device, highlighting each sentence and word as it is spoken.

Listen to any web page. Narrator of Mordor reads articles aloud — or just the text you select — and highlights each sentence and word as it is spoken, so you can follow along, listen while you do something else, or speed-read. PRIVATE BY DESIGN Speech is generated on your own computer by open text-to-speech models. Nothing you read is sent to a server, and there are no accounts, subscriptions or API keys. FEATURES • Reads the whole page, the detected main article, or your selection • Highlights the current sentence and word in colours that fit the site • Play/pause, back to paragraph start, rewind, next sentence, next paragraph and a draggable progress bar • Speed from 0.5× to 4×, with a words-per-minute estimate • Natural narration: pauses between paragraphs and after headings • Two voice engines: Kokoro (27 English voices) and Supertonic (31 languages, follows the page's language) • Voice Builder: create your own Supertonic voices by mixing and shaping the built-in ones • Speed reading (RSVP): shows the current word with a fixed focus letter, with or without the voice • Four colour themes — Mordor, Shire, Rivendell and Gondor — in light and dark • Keyboard shortcut Alt+Shift+N; the bar stays open per tab across reloads GETTING STARTED Click the toolbar icon (or press Alt+Shift+N) to show the bar, then press play. The first time a voice engine is used, its model is downloaded once (about 340 MB for Kokoro, 400 MB for Supertonic) and kept on your device. WebGPU is used when available; otherwise the voice runs on the CPU, which is slower. PRE-RELEASE This is an early version. Use the feedback button in the bar to report a bug or suggest a feature. Open source (GPL-3.0): https://github.com/chunky-bits/narrator-of-mordor Voice models: Kokoro-82M (Apache-2.0) and Supertonic 3 (OpenRAIL-M).