Safari is not supported. The voice model fails to load in Safari. Please try Chrome, Edge, or Arc for the best experience, or Firefox as an alternative.
Mobile browsers work, but this tool isn't optimised for them.
The voice model (~90 MB) must download before playback can begin — expect a wait on a mobile connection.
Audio generation is CPU-only on mobile, so it's slower than desktop and will drain your battery during long sessions.
For the best experience, use Chrome or Edge on a desktop or laptop.
Manuscript Reader.
Hear your writing read aloud: free, private, no AI training, no signup.
Free tool, running entirely on your device, so audio quality and speed depend on your hardware. Works best on a Mac in Chrome. That's the price of free. If generation is slow or the voice sounds garbled, your hardware likely can't keep up with the Kokoro model; try a different computer or browser.
Voices are generated by Kokoro, an open-source text-to-speech model trained on permissively licensed audio. We didn't build Kokoro, we just use the open-source library.
No servers. Your text is never uploaded anywhere. Everything runs inside your browser on your device.
✓
No AI training. Your words are not used to train any model. The voice model is pre-trained and read-only.
✓
No accounts or personal tracking. No logins, no cookies. We count page visits but don't know who you are or what you paste.
✓
Your text is never stored. Close the tab and your manuscript is gone. If you choose to "Remember my settings," only your voice and timing preferences are saved in your browser — never your text.
Best experience: Chrome or Edge on desktop
✓
Chrome, Edge, or Arc on desktop support WebGPU acceleration — audio generates 3-5× faster with smoother playback.
○
Safari is not currently supported — the voice model fails to load. Firefox works but uses slower CPU processing, so generation takes longer and may have brief pauses.
○
Mobile browsers work but are slower. For long chapters, desktop is recommended.
Loading Kokoro voice model...
First load downloads ~80-320 MB depending on browser (cached after)
This runs entirely on your device, so quality and speed depend on your hardware, and it works best on a Mac in Chrome. If generation is slow or the voice sounds garbled, try a different computer or browser.
Drop your file here
or (.txt, .docx)
Space play / pause · ← back 30s · → forward 30s
Applying changes will re-process the text — audio will regenerate from where you are now.
Voice
Speed1.0×
Sentence pause500ms
Paragraph pause1000ms
Speaker change pause1000ms
Action beat pause400ms
Comma pause250ms
Em dash pause250ms
Scene break pause2000ms
Saves your voice and timing preferences in this browser's local storage so they're restored next time you visit. Nothing is sent anywhere — it stays on your device.
Audio engine
Advanced. If your audio sounds glitched or distorted, try Force CPU. If you don't know what this is, leave it on Auto.
