Google Lens manga translation not working?
Why Google Lens often looks wrong on manga: vertical text, lettering on the art, one image at a time. What helps and when a manga translator fits.
Google Lens often looks wrong on manga because it is built to translate any image, such as a sign or a menu, not to rebuild a comic page. It places translated text over the original and has no idea what a speech bubble is. For a line or two it is fine, and for a chapter a tool that re-letters the page is easier to read.
Why Lens struggles with manga
- Vertical text. Japanese lettering is often vertical and split into short columns, and a translation of each column on its own can lose the sentence.
- Reading order. Bubbles are read in the order they are found, which is not always the order a human reads them.
- Lettering on the art. Sound effects and stylized titles are part of the drawing and are hard for a general text reader.
- Overlays. Translated text drawn on top of the old lettering can be hard to read on a busy page.
- One image at a time. There is no way to send a whole chapter in order.
Quick fixes that can help
- Use the biggest, sharpest copy of the page. Small or compressed images lose the thin strokes that tell characters apart.
- Zoom in on one bubble before you run Lens, so the text is larger on screen.
- Select the text region if your version allows it, instead of translating the whole image.
- Compare a doubtful line with a second reading before trusting it.
- Check that the source language is right, especially on pages with little text.
What a manga translator does differently
A manga translator treats the page as a page. It finds each speech bubble, reads the text, translates it, erases the old lettering and fills in the background, then sets the translation inside the bubble. You get the page back as an image in your language, for a whole chapter at a time.
The wording itself can still be similar on a short line. RuneManga translates the text with RuneLingua, its own AI translator, and Lens with Google's. The bigger difference is everything around the wording: the bubbles, the cleanup and the layout.
What to look for in a manga translator instead
- It finds speech bubbles, not just lines of text.
- It erases the old lettering instead of drawing on top of it.
- It takes a whole chapter in one go, in page order.
- It gives you the page back as an image you can keep.
- It tells you honestly what it cannot do yet.
These are the things Lens was never designed to do, and the reason a separate kind of tool exists for comics.
Use RuneManga instead of Lens
- Open the WebApp and add the pages: choose files or a folder, drop them in, or paste an image.
- Choose Translate.
- Set the original language, or leave Auto-detect, and pick the language you read.
- Press the button and wait for the page-by-page progress.
- Compare Original and Result, then download one page or the whole chapter as a ZIP.
When Lens is still the better choice
- You only need to understand one line, quickly.
- You are reading on paper or at a screen and want a camera translation.
- You need a language that a manga translator does not offer.
- You do not want to upload anything to a manga-specific service.
RuneManga is in beta and has gaps of its own. Text outside bubbles and sound effects are not translated yet and stay as drawn, and Korean and Chinese recognition is still improving.
Questions
Why does Google Lens translate manga badly?
Lens is a general image translator. It does not know speech bubbles, reading order or vertical comic lettering, so its output on manga can be incomplete or out of order.
Can Google Lens translate a whole chapter?
Not as a batch. You translate one image at a time.
Is RuneManga better than Google Lens?
They do different jobs. Lens is faster for one line, and RuneManga gives you back a re-lettered page for a whole chapter. See the comparison page for the details.
Which translator does RuneManga use?
RuneLingua, our own AI translator, built for manga and manhwa dialogue. It does the wording on every plan, Free included. Only the text goes to the translator, not the image.