Remove Hearing-Impaired Text from SRT Subtitles

How to remove hearing-impaired text from an SRT

1. Drop the subtitle file into the box. SRT and VTT are the usual inputs; ASS, SBV, TTML and the other text formats the site reads work too, up to 20 MB. The file card shows how many subtitles were read and when the last one ends.
2. Look at the five switches. All are on by default: square brackets, parentheses, music-only lines, speaker names and leftover dashes.
3. Check the preview of the first six cleaned subtitles. Untick a switch and the preview updates straight away.
4. Choose SRT or WebVTT under "Save as" and press "Clean subtitles and download". The file is saved with -no-sdh added to its name.
5. Read the message under the preview: it says how many subtitles were removed because they only described sounds, and how many more were cleaned.
6. Spot-check a few scenes further into the file, since the preview only shows the opening.

SDH, HI and CC files

SDH stands for subtitles for the deaf and hard of hearing. Alongside the dialogue they carry what the soundtrack tells a hearing viewer: sound effects, music, tone of voice and who is speaking when it isn't obvious. Files labelled HI or CC on subtitle sites are usually the same kind of thing. Current streaming style guides tend to put sound effects and speaker IDs in square brackets and mark sung lines with ♪, while older and fan-made files often use parentheses and capitalised names followed by a colon. The switches below are built around those habits.

If you need the opposite, adding sound cues and speaker names to a dialogue track, see creating SDH subtitles and the guide to captioning sound effects and music.

What each switch removes

Remove [sound descriptions]

Any text in square brackets or curly braces, anywhere in the subtitle and even across a line break, such as [DOOR SLAMS], [in Spanish] or {laughing}

Remove (descriptions in parentheses)

Any text in round brackets, anywhere in the subtitle, such as (sighs) or (ON RADIO), whether it is in capitals or not

Remove ♪ music-only lines

A line that both starts and ends with ♪, ♫ or # is deleted; other ♪ and ♫ symbols are stripped and the words beside them kept

Remove speaker names (JOHN:)

A name, a colon and a space at the start of a line, after an optional dash. It catches capitalised names of up to about 30 characters, including accented capitals, digits and initials, and one- or two-word names in title case such as Maria: or Mary Jane:

Tidy leftover dialogue dashes

When cleaning leaves a subtitle with one line, the dash in front of it is removed

After the switches run, lines left empty or holding only punctuation, such as a lone dash, are dropped and double spaces are closed up. A subtitle with nothing left is deleted completely. Its time slot stays empty; the neighbouring subtitles are not stretched to fill it.

A before-and-after sample

Subtitle 14 before

Two lines: "- [DOOR SLAMS]" and "- MARIA: Who's there?"

Subtitle 14 after

One line, "Who's there?", at the same time as before

Subtitle 15 before

"♪ Down by the river ♪"

Subtitle 15 after

Deleted and counted as removed

Subtitle 16 before

"(laughs) You're kidding."

Subtitle 16 after

"You're kidding.", counted as cleaned

The subtitles that remain are numbered from 1 again in the saved file, so subtitle 16 above becomes number 15 once 15 is gone.

Edge cases worth checking

The rules are simple pattern matches, which makes them predictable but not clever. These are the cases to look for:

• Song lyrics you want to keep. Untick the music switch: lines with ♪ then stay exactly as they are.
• Brackets used for something other than sound, such as a translator's note or an on-screen sign written as [Sign: Pharmacy]. The bracket switch removes those too; untick it and delete the sound cues by hand.
• Ordinary asides in parentheses, such as "I'll take the bus (again)." The parentheses switch removes every pair, not just capitalised or whole-line ones, so dialogue like this loses words.
• A single capitalised word before a colon. "Look: we don't have time." reads like a speaker label to the rule and becomes "we don't have time."
• Names the rule misses: names in Cyrillic, Greek, Arabic or East Asian scripts, mixed-case names like McCoy, titles such as Dr. Smith in title case, three-word names, and labels with no space after the colon.
• Filler words such as "uh" and "hmm" and lines in capitals for shouting are not touched. Desktop editors such as Subtitle Edit offer a more configurable remover, with interjection lists and more options, if you need that depth.

What is kept and what is lost

• Kept: every word of dialogue the switches don't match, and the start and end time of every remaining subtitle.
• Removed: subtitles that held only sound descriptions or music, and the matched parts of the others.
• Lost along the way: italics, colours and positioning, because the file is read as plain text, plus WebVTT cue settings.
• Not saved separately: the removed SDH text isn't written to its own file, so keep your original if you might want the full version later.

Limits

• One file at a time, up to 20 MB, text subtitles only. Image-based tracks from Blu-ray (PGS) or DVD (VobSub) need OCR first.
• The preview shows the first six subtitles; the count after saving covers the whole file.
• Output is SRT or WebVTT.
• The tool removes text. It doesn't rewrite or translate what is left.

Where to go next

• SDH subtitles still inside an MKV or MP4: save the text track first with the subtitle extractor, then clean it here.
• Accented letters showing as é or gibberish: re-save the file with the subtitle encoding converter before cleaning, so the name and bracket rules see the right characters.
• Dialogue-only subtitles burned into the picture: open the video and the cleaned SRT in the free subtitle editor, check the lines, style them and export an MP4. Rendering needs a desktop browser.
• Just the words: turning an SRT into plain text covers stripping numbers and timecodes as well.