Script
Used for your file names, for example my-narration.wav
Saved in this browser only, so it’s still here if you reload.
0 / 5,000 characters
How bracket cues work
Put directions in square brackets. They are never read aloud. A cue changes everything after it until the next cue. Use [normal] to go back to your usual voice.
| Type this | What happens |
|---|---|
[pause] [pause 1 second] [pause 2.5 seconds] | Silence at that point. Plain [pause] is 0.8 seconds; the longest is 10 seconds. |
[slowly] [very slowly] [quickly] | Changes the speaking speed. |
[whisper] [whispers] | Quieter and a little slower. Approximation: it is not a real whisper. |
[shout] [shouts] [shouts in an angry voice] | Louder, protected from distortion. Approximation: it is not a real shout. |
[angry] [sad] [excited] | Small changes to speed and volume. Approximation: the voice can’t really act emotions. |
[normal] | Back to your normal speed and volume. |
Any other bracketed text, like [laughs], is skipped and you’ll see a warning. A blank line between paragraphs adds a short natural gap.
Pronunciation fixes
If the voice says a name or word wrongly, spell it the way it sounds. Every time the word appears in your script, the voice will say your spelling instead. Your script text doesn’t change.
Split words into sounds with hyphens, like shi-vawn or nee-am. Try a few spellings and press play to hear each one.
Tips for natural-sounding narration
- Punctuation shapes the voice. A comma adds a short breath, and a full stop gives a proper ending. A question mark lifts the end of a question.
- Keep sentences short. Aim for under about 25 words. Long sentences tend to sound flat.
- Use dashes and ellipses for drama. “And then — silence.” or “Wait for it…” add a natural hesitation.
- Write numbers how you’d say them if they come out oddly: “nineteen sixty-six” rather than “1966”, “twenty-five percent” rather than “25%”.
- Spell out abbreviations that should be read as words, and use full stops between letters you want read separately: “N.H.S.”
- Leave a blank line between paragraphs for a natural gap, and use
[pause]where you need a longer one. - Try a couple of voices and speeds. A slightly slower speed (0.9×) often sounds calmer and clearer for narration.
Voice
Kokoro can change speed, but not pitch or emotion, so Voice Studio doesn’t offer those controls.
Generate
The voice model (about 92 MB) downloads the first time you generate, then it’s saved in this browser.
- Load model
- Generate speech
- Prepare audio
Waiting…
Your narration
You’ve changed the script or settings since this was made. Generate again to update it.
WAV is best for editing in Audacity. MP3 is smaller and works well in Canva. Both are mono, 44.1 kHz.