Why can’t Yomitan look up burned-in anime subtitles?
YuzuLingo · Nov 1, 2026
Yomitan looks up text that already exists in the page as selectable characters. Burned-in anime subtitles are painted into the video image, so the browser has no string for Yomitan to scan until something turns those pixels into text.
Softsubs versus hardsubs
A soft subtitle is a separate track or DOM overlay: SRT, ASS, WebVTT, or site-rendered captions. Players and extensions can expose that text to the page. A hardsub (burned-in) line is part of each video frame. There is no caption file and no text node—only colored pixels where the glyphs were drawn.
What Yomitan actually reads
Yomitan and similar hover dictionaries scan document text under the pointer, then match against installed dictionary zips. They do not open the video decoder, sample frames, or run OCR. If the subtitle never becomes HTML text, lookup has nothing to tokenize.
The OCR pipeline that fixes it
Hardsub mining tools crop the subtitle region, detect when the line changes, run OCR on that crop, and draw the recognized string as real overlay text. Once the line is text in the page, Yomitan, 10ten, Zhongwen, or a built-in popup can hover-scan it like any other caption. Offline OCR keeps the frames on the device; cloud OCR sends images out of process.
Why offline players still feel different
Desktop stacks such as mpv plus a mining script often read soft subtitle tracks or OCR a local window. Browser streams of anime and C-dramas frequently ship hardsubs only. The missing step is the same either way: obtain timed text, then look words up. Batch tools OCR a whole file to SRT for subs2srs; interactive tools OCR the live player so you never leave the video.
What still will not work
Installing more Yomitan zips does not create text from pixels. Blurring or covering the hardsub without OCR also leaves nothing to look up. You need either a real subtitle track or an OCR overlay that emits characters into the page before any dictionary extension can help.
