Free online tool — no sign-up
Upload an English EPUB and get the first fifteen pages back in Hebrew. Hebrew asks a question of every line of dialogue that English never answers: who is being spoken to, and what gender are they?
The first 15 pages of text are sent to our server to be translated. The file itself stays on your device.
A Hebrew verb agrees with the gender of the person it is addressed to. "You said" has two forms, and choosing between them is not optional — it is the verb. English gives a translator nothing to choose with, so the answer has to come from knowing who is in the room, which a paragraph in isolation does not tell you.
That is what reading the whole book first is for. The cast, their genders, their relationships, the recurring terms and the names go into a guide before any translating starts, and every page of the sample is written against it.
The other half of the job is mechanical and is where most tools stop: the reading direction has to be written into the file. Unicode will put the letters in the right order on its own, so untagged Hebrew looks passable and is still aligned to the wrong edge with its punctuation on the wrong side.
A DRM-free EPUB. The file stays on your device; only the text of the sample is sent.
The cast and their relationships are mapped first, which is what makes gendered verb forms come out right.
About 4,500 words of modern Hebrew, translated as continuous prose.
A complete EPUB with dir="rtl" written onto the translated text, so it aligns and punctuates correctly.
No nikud. Vowel points are omitted, which is how adult Hebrew prose is published. A book for learners would want them and this is not that.
Gender is inferred, not given. The English never stated it. The guide makes the inference consistent across the sample; it cannot make it authoritative.
Modern Hebrew, not biblical register. The translation will not imitate Tanakh style to match an archaic English.
Fifteen pages, not the book, and the text is sent to a server and discarded when the job ends.
Yes. The translated text carries an explicit direction attribute and a Hebrew language tag, so a reader aligns it to the right edge and places the punctuation correctly — which is exactly what text relying on Unicode ordering alone gets wrong.
From the cast mapped across the whole book. This is the clearest reason the model reads the entire text before translating: English "you said" carries no gender, and guessing per paragraph produces a character whose gender changes between chapters.
No. Adult Hebrew is published without them, and adding them would mean committing to one reading of every ambiguous word — which is a decision for a learner's edition, not a novel.
About 4,500 words, roughly fifteen pages, five books a day per person.
All free AI book tools · Free converters that run in your browser
The app runs the same engine across the entire book and can show the Hebrew next to the English, with your position synced between devices.
Open DualBook