Support

Speak selection reads the text you've highlighted with the mouse. Select a passage and try again, or use Read page to read the page's main article.

Your selection is inside a frame embedded from another site (an artifact or embedded document, for example). The extension can only read that frame with your permission: click the button and Chrome asks you to allow that one site. You can remove the access later in the extension's details in chrome://extensions.

Chrome keeps extensions out of some pages, such as chrome:// pages and the Chrome Web Store. Kokoro Speak works on ordinary web pages.

Open the extension's options. Under Pronunciation, add the word to Your pronunciations with how it should be said, or add an acronym to the list to spell out letter by letter (for example AWS, read as "A-W-S"). When writing letters yourself, join them with hyphens: "A-W-S", not "A W S".

The speech model loads when you first read something, which takes a few seconds. After that, speech starts quickly. It runs on the graphics card when Chrome can use it, and otherwise on the processor, which is slower. You can choose under Run the model on in the options.

Another extension or app may be using it. Open chrome://extensions/shortcuts to see or change Kokoro Speak's shortcuts.

It's optional and off by default. In the options, under AI rewriting, enter your own Anthropic API key and click Test to check it. Then only sentences with numbers or symbols are sent to Anthropic to be rewritten for speech. See the privacy policy.