Speak It

Read web pages aloud with the Kokoro text-to-speech model, running in the browser.

Speak It reads web pages aloud in a natural voice, right in your browser. Nothing you read is sent to a server: the speech is generated on your own computer by Kokoro, an open text-to-speech model. Open an article and click Read page. Speak It finds the main text and skips menus, ads and sidebars. It highlights each sentence as it reads and marks the word being spoken, and the page scrolls along with it. WHAT YOU CAN DO - Read the whole page, or just the text you've selected. - Right-click a paragraph and choose "Read page from here" to start in the middle. - Click any sentence while it's reading to jump to it. - Pause, skip back and forth between passages, and change the speed from 0.5× to 3× without the voice changing pitch. - Scroll away and the "Back to text" button takes you back to where it's reading. - Use the keyboard: Alt+Shift+R reads the selection (or the page), Alt+Shift+P pauses and resumes. VOICES 41 voices in American and British English, Spanish, French, Hindi, Italian and Brazilian Portuguese. WORKS WHERE YOU READ News sites, blogs and documentation, Google Docs (read from the document's text, without highlighting), and pull requests and issues on GitHub, GitLab and Forgejo, including their comments. PRIVATE BY DESIGN The text you read and the audio never leave your computer. There are no accounts, analytics or ads. The only download is the speech model itself (about 325 MB), fetched once the first time you open Speak It and kept on your computer, so later reads start quickly and work offline. GOOD TO KNOW Speak It is fastest on computers where Chrome supports WebGPU, which covers most recent ones. Without it, speech still works, but takes longer to start.