NoaLingua for Chrome
Add to Chrome — Free Add to Chrome

Feature

The whole script of a video, as a file, with an honest count

The transcript panel shows a video’s entire script rebuilt into sentences, with search, a difficulty estimate and a list of words you have not met. You can export it as plain text instantly, or as a bilingual PDF that translates every sentence and then reports exactly how many lines succeeded.

How to use it

Step by step

  1. 1

    Open the transcript from the quick menu

    It docks into YouTube’s own sidebar above the recommendations, and floats over the player in fullscreen.

  2. 2

    Search it, or read it as running text

    Search jumps between hits. “Read as text” switches to prose, with every word clickable exactly as in a subtitle.

  3. 3

    Check the difficulty and the new words

    Difficulty is computed across every word spoken, compared against your deck. New words are ordered by how often they occur in this video.

  4. 4

    Export as .txt or as a bilingual PDF

    Plain text is instant and untranslated. The PDF translates every sentence of the video — not just the part you watched — and lays original and translation one under the other.

Mechanism

Why it behaves this way

Why the export message counts lines

A bilingual PDF can fail partially for several ordinary reasons: a cloud translation quota running out mid-document, a language pair with no model, a model still downloading, or a translator that simply stops answering.

Rather than producing a document with silent gaps, the finished message states the actual number: “Downloaded — 412 of 741 lines translated,” with the reason. You then know whether to wait for a quota reset, download a local model, or accept the partial file.

Why pressing the button is what allows the download

Chrome requires a genuine user gesture before it will let a page start downloading a translation model. Pressing Export is that gesture, which is why the export sequence begins by preparing the translator — and why that wait is capped at ten seconds so the export cannot freeze.

Punctuation for captions that have none

Auto-generated captions frequently have neither full stops nor commas. Where the punctuation model is downloaded and more than half the transcript’s sentences lack punctuation, it is restored — but only if the model returns exactly the same words. Change one word and the original stands.

Full stops are never inserted merely because a speaker paused. That would be inventing sentence boundaries rather than restoring them.

Limitations

Common questions

Does the PDF cover the whole video or only what I watched?

The whole video. Sentences you already watched come from cache and cost nothing; the rest are translated during the export.

What is in the plain text export?

The original script with timestamps, the video title and the URL. It is instant because nothing is translated.

What does the difficulty estimate actually measure?

How hard the video is for you specifically — it compares every word spoken against your own deck, rather than reporting a general reading level.