Natural-sounding speech, without the cloud
Spoken replaces Chrome's built-in Speak with high-quality AI voices. Every word is synthesized on your own GPU, so nothing you read leaves the browser.
What Spoken does
Zero cloud dependency
All inference runs locally in the browser. Your text is never uploaded.
No API costs
WebGPU does the synthesis, so there is no per-character bill.
Near-native quality
Kokoro voices give clear, natural speech instead of robotic system voices.
Instant after load
Once the model is cached, playback starts without waiting on a network round trip.
Works without WebGPU
The voice model downloads and caches automatically, and falls back to WASM when WebGPU isn't available.
Playback controls
Speed, play, pause, stop, a progress bar and keyboard shortcuts.
How it works
Select text
Highlight anything on a page: an article, a doc, an email.
Right-click, Spoken: Speak
The context menu entry starts reading straight away.
Listen locally
Pick a voice and speed. Audio is generated on your machine.
- < 2s
- Time to first audio (target)
- < 5s
- Full synthesis of 100 words (target)
- < 10s
- Model load time (target)
- < 500 MB
- GPU memory (target)
Questions
Try Spoken
AI text-to-speech for Chrome that runs entirely in your browser on WebGPU.