Copy the raw Japanese, Korean or Chinese out of a comic page - in the order a human would read it, not the order a scanner finds it. Vertical text and speech balloons included.
JPG, PNG, WEBP - text comes back in reading order, right-to-left for manga
Balloons are ordered right-to-left for manga and top-down for webtoons. A generic OCR hands you scrambled fragments.
Tategaki is read as columns, not as a stack of unrelated single characters - the failure mode of every general-purpose OCR.
The same detector the translator uses tells dialogue apart from sound effects and from text printed on the artwork.
Each line comes back with a translation next to it, so you can check the OCR without reaching for a second tool.
Nothing to install and nothing to sign up for.
Yes. The translator detects the balloons, reads the text and writes the translation back into the art. OCR mode stops after the reading step and hands you the text.
Yes, and that is the main reason it exists. General OCR reads vertical text as a column of unrelated characters; here the balloon detector groups it first.
Yes. Copy all text gives you the source text only, one line per balloon, in reading order.
Hand-lettered text and stylised SFX are the hardest case for any OCR. Printed dialogue is reliable; expect misses on heavy handwriting.