"Is this book too hard for me?" is a measurable question.
Drop the file in. It never leaves your laptop.
Give it any EPUB and you get a reading grade level, a reading-ease score,
vocabulary richness, and the chart that actually matters to a learner: the
coverage curve showing how few distinct words you need to
follow most of the book. The EPUB is unzipped
in the browser with a native API — no server, no upload, and
no third-party libraries.
A real run on the bundled public-domain sample.
Alice's Adventures in Wonderland: Grade 6,
26,715 words, and 1,397 distinct words cover 95% of it.
0 bytes
uploaded. The ZIP is decompressed by the browser itself, so the book
never touches a network.
0 deps
No unzip library, no charting library, no framework. One HTML file.
1,397
distinct words for 95% coverage of Alice. The top 100 alone
carry 59% of the book.
1 click
to try it — a public-domain sample is bundled so you don't need a file
to hand.
↓ the trick
An EPUB is a ZIP, and browsers can now unzip
Every EPUB is a ZIP archive of HTML files. The usual way to read one in a web
page is to pull in a JavaScript unzip library, and the usual way to analyse a
book is to send it to a server. Both are avoidable.
This uses DecompressionStream,
a native browser API, to decompress the archive in place. The text is then
extracted with the browser's own HTML parser, tokenized and scored. The
consequence is the privacy claim: there is no upload path in the code
at all, so the promise is architectural rather than a policy you have
to believe.
Reading grade
Flesch–Kincaid grade level, plus a 0–100 reading-ease score.
Vocabulary richness
Type–token ratio, with total and unique word counts alongside it.
Top words
The twenty most frequent meaningful words once stopwords are removed.
Coverage curve
How much of the book you can read knowing the N most common words. The
most useful chart for a learner.
↓ why the coverage curve
The number a learner actually wants
A grade level compresses a whole book into one figure and tells you almost
nothing actionable. The coverage curve answers the real question: how much
vocabulary do I need before this stops being exhausting?
For Alice, the answer is 1,397 distinct words for 95%.
The top 100 words alone already carry 59%. That shape —
steep at first, then a long flat tail — is the same shape I measured across
a season of television in the
Love Island lexical analysis,
which is where this tool came from. That study needed a Python pipeline; this
turns the same lens into something anyone can point at their own book.
The whole tool, top to bottom.
↓ before you drop a file
It works on a phone too
Empty state on the left, phone layout on the right. The drop zone doubles as
a file picker on touch devices, and the bundled sample means the tool is
usable in one tap without hunting for an EPUB first.
Empty state — drop, pick, or try the sample.390px — cards stack, charts stay readable.
↓ what this cannot tell you
Readability formulas are cruder than they look
Flesch–Kincaid is essentially sentence length and syllable count. It has no
idea whether the ideas are hard, whether the vocabulary is archaic, or whether
the sentences are doing something structurally difficult. A book of short
sentences about difficult things scores as easy.
Type–token ratio is also sensitive to length: longer texts tend to score
lower simply because they have more running words. Comparing TTR between a
novella and a doorstop is not comparing like with like, and the tool does not
correct for it.
The coverage curve is the most robust thing here and even it counts word
forms rather than word families, so "run", "running" and "ran" are three
words. That inflates the count somewhat against the word-family figures used
in language-teaching research.
Finish · what it is built on
Deliberately boring technology
Unzipping
Native DecompressionStream, no third-party library
Text extraction
The browser's own HTML parser over the archive's XHTML files
Scores
Flesch–Kincaid grade and reading ease; type–token ratio for richness
Charts
Drawn on <canvas>
Accepts
.epub and plain text
Sample
Alice's Adventures in Wonderland, Project Gutenberg, public domain
Privacy
No server, no upload, no analytics. There is no code path that sends the file anywhere
