Browser extension · Arabic-first

Manga in Arabic,
the way it should be read

Qaari translates manga, manhwa and manhua in place on the page you are already reading. Letters join, text runs right to left, and the translation stays live text — never pixels burned into the image.

Live render — try selecting the text yourself

It isn't over yet

Don't open that door

Will we make it in time?

This is real text on the page. Select it. Copy it. Restyle it.

What you get

Everything you need to read manga in Arabic

No setup, no steps. Open a chapter, press one button, read.

01

Read where you are

The translation appears over the page you are already reading. No uploads, no downloads, no switching tabs.

02

Arabic done right

Letters join, text runs right to left — the way Arabic is actually written, not scattered characters.

03

Text you can select

The translation is real text: select it, copy it, look up any word you don't know.

04

Names that hold

A character is spelled the same way in chapter 1 and chapter 400, however long the series runs.

05

Your font, your size

Make the text bigger or change the typeface any time — instantly, with no re-translation.

06

Clean artwork

The translation replaces the text inside the balloon only, without painting over the art around it.

07

Manga, manhwa, webtoons

Japanese, Korean and Chinese pages — and long vertical scrolling strips too.

08

Clear about what is sent

Detection and reading happen on your machine. Only translation sends the text and a reduced copy of the page, and we keep neither.

The difference

Arabic you can actually read

Run any other tool on an Arabic page and you see it immediately: letters come apart, words sit backwards. With Qaari you read an ordinary Arabic sentence — the way a human translator would have written it.

Other tools

ل‌م‌ ‌ي‌ن‌ت‌ه‌ِ‌ ‌ا‌ل‌أ‌م‌ر‌ ‌ب‌ع‌د

Disconnected letters · reversed order

Qaari

لم ينتهِ الأمر بعد

Joined letters · correct direction

This is not a small technical detail — it is the difference between a page you read and a page you decipher.

The core difference

Live text, not pixels

Every other tool in this category burns the translation into the image. At that moment the translation becomes a picture: it cannot be selected, cannot be copied, cannot be read by a screen reader, and its font cannot change without re-running the entire pipeline.

Select and copy

The translation is actual text on the page. Drag across it like any other text.

Accessible

A screen reader reads the translation aloud. A burned-in image is silent.

Restyle instantly

Size, family and weight change live — no re-detection, no re-translation.

Refits itself

Binary search for the largest font size that fits the balloon, on both axes.

Try a font

It isn't over yet

Don't open that door

Will we make it in time?

Consistency

Names that hold across chapters

Every other tool translates a page in isolation. So a character is spelled one way in chapter 1 and another way in chapter 40. Qaari keeps a per-series glossary and carries it into every request.

Without a glossary
Chapter 1Kaito
Chapter 12Kaitou
Chapter 27Kaitoh
Chapter 40Kayto
Qaari · series glossary
Chapter 1Kaito
Chapter 12Kaito
Chapter 27Kaito
Chapter 40Kaito
An established rendering is never rewritten. Once a name has a form, later chapters reuse it even when the model would now word it differently.
Comparison

Where Qaari actually differs

The category is crowded and the hard parts are mature open source. These are the differences that survive testing.

QaariOther tools
Translation stays live textonly
Yes — selectable, copyable, screen-reader readable
No — burned into the image
Name consistency across chaptersonly
Per-series glossary, established forms locked
Each page translated in isolation
Change the font after translatingonly
Instant, no reprocessing
Requires a full re-run
Arabic shaping and direction
Browser engine — correct by default
Mixed; some handle it, many do not
Pages leave your machine
Only extracted text is sent to translate
Usually the full image is uploaded
Accuracy stated with a method
18/18 across three pages, gaps published
Marketing figures with no methodology

And where it does not win — yet

OCR on CPU takes about 13 seconds per page and needs a local server running. Cloud tools start faster and install more easily. Recall on hand-drawn sound effects is still poor. We would rather say so now than have you find out after paying.

Privacy

Exactly what gets sent

Balloon detection, text recognition, erasure and rendering all run on your machine. Only translation sends anything: the extracted text and a reduced copy of the page (1024px on its longest edge), because tone and speaker cannot be recovered from words alone. Neither is stored once the request finishes.

  • Host access is requested per site at the moment of use, never blanket
  • The glossary lives in your browser, not on our servers
  • Default fonts are OFL-licensed or system-provided
Launch

Qaari has not launched yet

The engine works and the Arabic is proven. What remains is accounts, billing and shipping the extension. Leave your email and you will be among the first to try it.

No marketing. Launch notice only.

FAQ

Questions you should be asking

It works on pages that serve images directly. Host access is requested per domain the first time you use it there, rather than a blanket grant across all sites — which is also the single most common reason extensions of this kind stall in Chrome Web Store review.