Extract subtitles from an MKV or MP4

Subtitles inside a video come in two kinds. Text subtitles store words and times, so they can be saved as a subtitle file. Picture subtitles store a small image of each line, exactly as it appears on screen, so there are no words to save without character recognition.

If you can see subtitles while the video plays but no subtitle track is listed, the words are probably burned into the picture, so they are part of the video frames and cannot be extracted. Captions carried inside the video stream itself, as some TV recordings do, are not a separate subtitle track either.

Films often carry several subtitle tracks in the same language. The labels in Tracks tell them apart:

• Forced tracks contain only the lines meant to show even with subtitles off: dialogue in another language, signs, a letter on screen. They are short, often a few dozen lines.
• Default marks the track a player turns on first. It is often not the one you want.
• The name usually says what the track is: “English”, “English SDH” (with sound descriptions for deaf and hard-of-hearing viewers), “Commentary”, “Signs & Songs”.
• The language is a code such as eng, fre or jpn. A track marked und has no language set; check its name or open the downloaded file.

If the labels are wrong or missing, you can fix them in the MKV metadata editor so players pick the right track automatically next time.

Blu-ray (PGS) and DVD (VobSub) subtitles are sequences of images with timings. To turn them into SRT you need optical character recognition (OCR): software that reads the letters in each image. This tool does not do OCR, so it lists these tracks but cannot save them as text.

1. Extract the picture track as it is with MKVToolNix (below). PGS comes out as a .sup file, VobSub as an .idx and .sub pair.
2. Open that file in a subtitle editor with OCR, such as the free Subtitle Edit on Windows.
3. Run the OCR, correct the mistakes it makes (often with letters like l and I), and save as SRT.

Many players can show PGS and VobSub tracks directly, so if you only want to watch with subtitles, you may not need to convert them at all.

MKVToolNix

First list the tracks and their IDs, then extract by ID. IDs start at 0 and include the video and audio tracks:

              mkvmerge --identify film.mkv
mkvextract film.mkv tracks 2:film.en.srt 3:film.en.forced.srt 4:film.fr.sup
            

mkvextract writes each track in its own format: text subtitles as SRT, ASS as ASS, WebVTT as VTT, PGS as SUP and VobSub as IDX/SUB. It does not convert between them.

FFmpeg

              ffmpeg -i film.mkv -map 0:s:0 film.en.srt
ffmpeg -i film.mkv -map 0:s:1 -c copy film.signs.ass
ffmpeg -i clip.mp4 -map 0:s:0 clip.srt
            

0:s:0 means the first subtitle stream of the first input. Without -c copy, FFmpeg converts a text track to the format implied by the file extension, which is handy for MP4 timed text but drops ASS styling. FFmpeg cannot turn picture subtitles into text either. The ffprobe guide shows how to list streams first.

• Play it with the video: most desktop players, VLC included, load a subtitle file that sits next to the video with the same name, such as film.mkv and film.srt. A language code before the extension (film.en.srt) helps media servers label it.
• Edit or translate it: SRT is plain text with numbered blocks and times, so any text editor works. Keep the times untouched unless you mean to shift them.
• Keep it as ASS if the subtitles are styled (anime, karaoke, positioned signs). Converting ASS to SRT throws away fonts, colours and positions.

This page deals only with subtitles that already exist inside a file. If a video has no subtitle track at all, there is nothing here to extract.