What is the difference between hardcoded and soft subtitles?

Soft subtitles are a separate track in the file, like an SRT inside an MKV or MP4, which players switch on and off and tools can copy out exactly. Hardcoded subtitles were drawn into every frame when the video was made, so they can only be read back from the picture. If your video has a subtitle track, the Video Cutter can save it as a file without OCR.

Some subtitles were missed or split in two. Why?

A subtitle is missed when it falls outside the box, or when its text does not stand out from the picture in the chosen style; try a taller box or another text style. One that fades in or out, or a busy background, can make one subtitle look like two; repeated lines are joined automatically, and you can join others with the join button.

Can it remove the subtitles from the video?

No. Burned-in text replaced the picture underneath it, so it cannot be lifted off; this page only reads it. Cropping the strip off, or covering it, are the usual ways round it.

Is the video sent anywhere to be read?

No. The frames are decoded and read by your browser, on your computer. The only download is the OCR language data the first time a language is used.

Stays on your computer Downloads the OCR language data (a few megabytes) the first time each language is usedThe language data comes from tessdata.projectnaptha.com, which learns the language you chose, not the videoHow this works

Extract hardcoded subtitles from a video

Turn subtitles burned into a video's picture into an SRT or VTT file you can edit: mark where they appear, and each one is read with OCR and timed. In your browser, with nothing uploaded

Drag & drop a video with burned-in subtitles here, or browse

MP4, MOV, MKV or WebM · nothing is uploaded

About extracting hardcoded subtitles

Hardcoded (burned-in) subtitles are part of the picture itself, not a track that can be switched off or copied out, so the only way to get them back as text is to read them like a person would. This page looks at the strip of the picture where the subtitles sit, notices each time the text changes, reads each subtitle once with OCR (optical character recognition), and gives you the lines with their times as an SRT or WebVTT file. It is free to use, with no account and no upload limit, and it runs entirely in your browser: nothing is uploaded, and while the page stays open, a feature you have already used keeps working offline. The OCR language data (a few megabytes) is downloaded the first time a language is used and then kept by your browser.

What you can do

Mark the area the subtitles appear in, at the bottom, the top or anywhere else
Read white, yellow or dark subtitle text in 13 languages, including Chinese, Japanese, Korean, Arabic and Russian
See each subtitle's picture next to its text, to check and correct it
Edit the text and times, delete false finds, and join lines that were split
Save the result as an .srt or .vtt subtitle file

How to extract burned-in subtitles

Drop the video onto the page.
Move to a moment where a subtitle shows and draw a box around it, leaving room for two lines.
Choose how the subtitles look and their language, then press Find and read the subtitles.
Check each line against its picture, correct any mistakes, and save the SRT or VTT file.

How accurate is it?

Clear subtitles in a plain font, white with a black outline, are usually read almost word for word. Mistakes come from small or blurry text in a low-resolution video, fancy or italic fonts, text over a bright or busy background, and karaoke-style lines that change colour as they are sung. Common slips are similar-looking characters ("l", "I" and "1"; "rn" and "m") and missing punctuation. Each subtitle is first found at the checks per second you choose, then timed to the exact frame it appears and disappears on by a second pass through the video; without that pass, times are as precise as the checks (a quarter of a second at 4 a second).

Questions

What is the difference between hardcoded and soft subtitles?: Soft subtitles are a separate track in the file, like an SRT inside an MKV or MP4, which players switch on and off and tools can copy out exactly. Hardcoded subtitles were drawn into every frame when the video was made, so they can only be read back from the picture. If your video has a subtitle track, the Video Cutter can save it as a file without OCR.
Some subtitles were missed or split in two. Why?: A subtitle is missed when it falls outside the box, or when its text does not stand out from the picture in the chosen style; try a taller box or another text style. One that fades in or out, or a busy background, can make one subtitle look like two; repeated lines are joined automatically, and you can join others with the join button.
Can it remove the subtitles from the video?: No. Burned-in text replaced the picture underneath it, so it cannot be lifted off; this page only reads it. Cropping the strip off, or covering it, are the usual ways round it.
Is the video sent anywhere to be read?: No. The frames are decoded and read by your browser, on your computer. The only download is the OCR language data the first time a language is used.

To fix the timing of the saved file, use the Subtitle Fixer; to add it to a video as captions, Add Subtitles to Video; to read text in photos and scanned pages, the OCR Toolkit.