About

Speak naturally.
It follows you.

WordTrail is a real-time script reader that highlights words as you say them. Load in a script, hit record, and the active word stays centered on screen so you never lose your place.

Why WordTrail exists

Reading from a script is harder than it looks. Whether you're rehearsing a wedding speech, recording a podcast, narrating a video, or running through presentation notes, the moment you glance away or stumble over a word, you lose your place. Conventional teleprompter apps make you scroll at a fixed pace and force you to keep up. That's backward — the script should keep up with you, not the other way around.

WordTrail listens to your voice in real time and tracks where you are in the script. When you slow down, it slows down. When you skip ahead, it catches up. When the speech recognizer mishears a word, the matcher recovers gracefully and keeps the highlight where it should be.

How it works

The matching engine has two stages, both running over a sliding window around the last matched word:

  • Phrase matching on significant words. Common stop words like a, the, and, to are filtered out of both your spoken transcript and the script. The matcher then searches for the longest phrase from your recent speech that appears in the script, scoring by phrase length and proximity to where you should be.
  • Sequence alignment fallback. If phrase matching finds nothing, a Needleman-Wunsch-style alignment runs against the window. Word similarity is character-based — prefix match, or 75%+ positional overlap — chosen for speed. The result is accepted only if it clears minimum match thresholds.

Highlighting uses the CSS Custom Highlight API, which means rich-text formatting in your scripts is preserved exactly as you authored it. Nothing in the DOM is mutated — the highlight is purely a paint-layer effect on top of the existing text.

What it's good for

  • Speeches and presentations — keynotes, weddings, eulogies, best-man toasts, conference talks, Toastmasters practice.
  • Video and podcast narration — read scripts naturally without forcing your pace to match a scrolling teleprompter.
  • Voiceover work — keep your eye on the next word, not on re-finding your place after a take.
  • Language practice — read aloud in any of 12 supported languages and watch your pronunciation map to text in real time.
  • Accessibility — for readers who benefit from word-level visual tracking while reading aloud.

Languages

WordTrail supports 12 languages: English, Simplified Chinese, Traditional Chinese, Spanish, French, German, Italian, Portuguese (Brazil), Japanese, Korean, Russian, and Dutch. CJK languages (Chinese, Japanese, Korean) are tokenized per-character because there are no word boundaries to lean on.

Privacy by design

WordTrail does not run a server. There is no account to create. Your scripts live in your browser's localStorage on your own device — nothing is uploaded. Microphone audio is processed by your browser's built-in speech recognition service; WordTrail only ever sees the recognized text and never the raw audio.

For full details, see our Privacy Policy .

Tech stack

WordTrail is a Progressive Web App built with:

WordTrail is open source under the MIT license. The full source code and matching algorithm are available on GitHub.

Browser support

Speech recognition through the Web Speech API works best in Chrome, Edge, and Safari. Firefox does not currently support the Web Speech API for recognition. For the most reliable microphone behavior on mobile, install WordTrail to your home screen as a PWA rather than running it in a browser tab.

Who built it

WordTrail is built and maintained by Daniel Kwok. If you want to suggest a feature, report a bug, or contribute, visit the GitHub repository or email kcfdaniel@gmail.com.