Release notes

What’s new in Readalix

Everything worth knowing about each release, newest first.

1.5.0 30 August 2026

New

  • Step back a sentence, or skip one: the left and right arrow keys move the read one sentence at a time, from the toolbar or straight from the page. How it works
  • Save a file in any voice that can make one: Piper, Kokoro, Gemini and OpenAI voices now write files too, instead of being swapped for a Microsoft voice. How it works
  • Chrome’s own “Google … (Natural)” voices are back: they are offered again once Readalix has quietly checked, this browser session, that they answer. How it works

Improved

  • Very long sentences are broken into readable pieces: text without full stops is cut at its commas and clauses, so the voice pauses and the tint stops covering half a paragraph. How it works
  • A saved file is named after what it says: the file now always waits for Chrome’s on-device model to title it, in any language, instead of sometimes falling back to the tab title. How it works
  • Every saved file credits the voice that read it: and the open-licensed voices carry their licence, with a note before you generate when that licence asks something of you. How it works
  • A run-on sentence turns a page in a video: instead of shrinking to fit one screen, it is dealt evenly over as many screens as it needs, at full size. How it works
  • The language and voice menus keep one order: rows no longer float to the top for being used recently, so a position stays learnable; drag them in Settings to set your own. How it works

Fixed

  • Pages whose content sits in one big frame did nothing: play now finds the article inside a large frame, so viewer shells and app pages read like any other page. How it works
  • Long Piper reads and files stopped partway: a long article or export in a Piper voice failed mid-way with an Aborted error; it now runs to the end whatever the length.
  • A stray “|” could end a read or fail a file: a separator on its own line in a byline made the voice return nothing, which stopped the read; such fragments are now skipped.
  • The voice that spoke could differ from the one named: on a site pinned to a language, the read used whichever voice was picked last anywhere, not the one the toolbar showed for that language. How it works
  • Videos in some voices had no word highlight: Piper, Gemini and OpenAI report no word timings, so their files came out with only the sentence colour; the word is now marked in every voice. How it works
1.4.0 28 August 2026

New

  • Save what you are reading as an MP3: a new button on the toolbar turns whatever Play would have read (the article it finds on the page, the passage you selected, a PDF, a Google Doc) into an audio file, so you can listen to it later, away from the browser. It is read by the Microsoft voices, which are the ones that can be written to a file.
  • …or as a video: the same text as an MP4 that shows one sentence at a time with the spoken word lit in your own highlight colours: something you can watch, post, or hand to someone who reads along. Pick it wide for a screen or tall for a phone.
    The export card: audio, tall video and wide video as three icons, switches for Downloads and Google Drive, and a Generate button.
    Pick audio or the video’s shape, choose where it lands, then Generate.
  • Send it to Google Drive: audio and video can go straight into a Readalix folder in your Drive instead of the downloads folder, or into both at once. Readalix only ever sees the files it puts there.
  • PDFs previewed in OneDrive and SharePoint read aloud: open a PDF from a OneDrive or SharePoint file list and press play: it reads the document, or the passage you selected in it, with the word and the sentence highlighted in your colours. No downloading it first.
  • A flag for every language: the language picker, the language settings and each language’s voice list now show the flag beside the name, so you can find the one you want at a glance. The toolbar’s language button can wear that flag too, next to the code or in place of it.
    The language picker, each language listed in its own script with its flag beside it.

Improved

  • Changing the language takes effect at once: picking a different language while something is being read now restarts the sentence you are on in that language, the way changing the voice or the speed already did, instead of waiting for the next read.
  • Settings opens at the size of what you opened: short sections such as Shortcuts or Toolbar no longer open as a mostly-empty window; the window is now as tall as the section needs.

Fixed

  • Reading could hang on “Preparing…” forever: certain voices offered in the list (Chrome’s own “Google … (Natural)” ones) are loaded only while Chrome’s reading mode is using them, so a voice you had picked could quietly be gone after a restart, and pressing play with it left the spinner turning with no way out but Stop. Those voices are no longer offered, and any voice that stays silent is now swapped for one that works so the reading carries on.
1.3.0 21 August 2026

New

  • ChatGPT, Gemini and Claude now show where they are reading: press read aloud on an answer and Readalix highlights the word and sentence their voice is speaking, scrolling the chat along with it. Their voice, Readalix’s highlighting, on every answer, and your reading speed applies to their voice too. Code blocks are left alone, because their voices skip them. You can switch any of the three off under Settings → Reading.
  • Press play and it finds the article: on an ordinary page with nothing selected, Readalix now works out where the article actually is and reads that, leaving the menus, sidebars and footer out. Nothing to select first, and the picker is still there when you want to choose a region yourself.
  • Turn the word or sentence highlight off: each half now has its own switch next to its colours, so you can follow just the word, just the sentence, or listen with nothing highlighted at all.

Improved

  • Reading options are easier to set: each option under Settings → Reading is now one block you can click anywhere on, description included, instead of only the title.

Fixed

  • Some PDFs refused to read at all: a PDF the site builds and opens for you, or one served from a download link or a path without .pdf in it, left the play button greyed out as if the page were off limits. Those all read and highlight now.
  • Reading didn’t scroll into view in chat-style pages: on sites where the page scrolls inside a panel, the reading no longer followed along and the jump-back chip did nothing when clicked. Both work there now.
  • Part of a selection went unread in chat-style pages: on sites where the page scrolls inside a panel rather than the window, anything you had already scrolled past was silently skipped. It now reads in full.
  • A line break no longer runs two sentences together: text split across a line break was read and highlighted as one long sentence, so the highlight covered more than you were hearing. Each part is its own sentence now.
1.2.0 18 August 2026

New

  • PDFs inside a page read too: a PDF shown within another page, the way a course site shows a handout or an invoice appears inline, now reads and highlights just like a PDF opened on its own. With several on one page, the one you select in is the one that reads, and clicking another stops it.
  • Gmail and Drive attachment previews read aloud: press play on a PDF you’re previewing in Gmail or Drive and it reads the document, or just the passage you selected in it, following along line by line. Scanned documents read too.

Fixed

  • Pause did nothing on most voices: with a Microsoft, Gemini, OpenAI or on-device voice, pressing pause left the reading going. Pause and resume now work on every voice.
  • Google Docs read the document twice: pressing play a second time could start two readings at once, in two different voices talking over each other.
  • Feeds and card grids ran on as one sentence: where a page stacks text in cards, like a subscriptions or search-results list, a whole column was treated as a single endless sentence. Each line is now its own sentence, so the highlight follows what you’re hearing.
1.1.0 15 August 2026

New

  • A Microsoft voice by default: reading starts on a Microsoft neural voice for the page’s language, so it sounds natural, with every word still highlighted exactly, before you pick anything. Pick any other voice and it’s remembered per language.
  • The accent next to every voice: voice names now say which country they speak for (Neerja · India, Sonia · United Kingdom), so you can pick the English (or Spanish, or Arabic…) you actually want. Search the voice list by country too.
  • Settings save themselves: every change applies the moment you make it, so there’s no Save button left to forget.

Improved

  • Local files explain themselves: opening a file from your computer used to say reading wasn’t supported there. It now tells you about Chrome’s file-access switch, which is all that was ever missing.

Fixed

  • The right-click menu named the wrong action: it was always one right-click behind, so it could offer to read a selection when nothing was selected.
  • A voice download showed 100% for a moment before starting again at 0%.
  • The popup stayed settings-sized after you left Settings instead of shrinking back to the toolbar.
1.0.0 6 August 2026 First release

The first public release of Readalix. Read any text aloud with the current sentence tinted and every word lit in sync with the voice.

Highlights

  • Read aloud with live highlighting: sentence by sentence, tinting the current sentence and lighting each word as it’s spoken. Both colours are customisable and stay readable on dark sites.
  • Works everywhere text lives: plain pages, form fields, shadow DOM, iframes, the built-in PDF viewers of Chrome and Edge, Google Docs, Google Slides, Word Online and PowerPoint Online.
  • Many voices: your browser’s own voices (free), Piper (100+ on-device voices in 30+ languages), Kokoro (28 on-device English voices), Microsoft neural voices (322, no key or download), plus Gemini and OpenAI cloud voices with your own API key.
  • Read without selecting: the whole document, or from the clicked word to the end, in Google Docs, Word Online and PDFs.
  • Customisable popup toolbar: play/pause, an element picker, and Language, Voice and Speed pickers. Drag to reorder, hide buttons, and auto-start reading or picking when the popup opens.
  • Live speed and voice changes: applied to the sentence being read, not only the next one.
  • Scroll-follow with a jump-back chip when you scroll away from the read.
  • Keyboard shortcut to start and stop. Even a single function key can be the shortcut.
  • Speech cache with a manager to review, replay and clear stored audio.
  • Settings sync across your devices; your API keys and downloaded voice models stay on the device.
  • 52 interface languages, listed by their native name and searchable by native name, English name or country.