How to Convert an SRT File to Plain Text

What you're removing

An SRT cue has three parts: a number on its own line, a timing line such as 00:02:14,500 --> 00:02:17,900, and one or two lines of text, followed by a blank line. Plain text keeps only the text lines. The structure is explained in what an SRT file is.

The harder part is readability. Subtitles break sentences across cues to suit reading speed, so a straight strip leaves a column of fragments. Joining them back into paragraphs is what makes the result usable.

With find-and-replace

Use an editor that supports regular expressions (VS Code, Notepad++, Sublime Text, BBEdit and many others):

1. Remove cue numbers: replace ^\d+\s*$ with nothing.
2. Remove timing lines: replace ^\d{2}:\d{2}:\d{2},\d{3} --> .*$ with nothing.
3. Collapse the empty lines left behind: replace runs of two or more line breaks with a single line break.
4. Join lines into running text: replace single line breaks with a space.
5. Re-insert paragraph breaks by hand at topic changes, or after every few sentences.

Check the result for doubled spaces and for formatting tags such as italic markers, which you can remove with another replacement.

With a short script

For many files, a script is quicker and repeatable. The logic in any language: read the file, skip lines that are only digits, skip lines containing the arrow -->, skip blank lines, join the remaining lines with spaces, and write the result. Add a rule to start a new paragraph when the gap between one cue's end and the next cue's start is longer than a couple of seconds; pauses often mark a change of thought, which gives you natural paragraphs for free.

Keep a few timestamps

Fully stripped text loses the link back to the recording. For research notes, quotes and review, keep a timestamp every few paragraphs or at each topic change, written in square brackets such as [12:40]. It makes checking a quote against the audio a matter of seconds. How to format a transcript covers timestamp styles and layout.

Start from a mydubly plain transcript

If you're making subtitles anyway, you may not need to convert. Each run of video to text on mydubly returns a plain transcript and a timestamped transcript alongside the SRT and VTT, from the same speech recognition. Upload a video (it stays on your device; only the audio is sent) or audio file and download whichever form you need, for 1 credit per minute. In another language, pick a target and the plain transcript comes back translated too. The timestamped transcript page compares the outputs.

Turning the text into something people read

Plain text from speech still reads like speech: repetitions, false starts and long run-on sentences. For an article or a summary, edit for the reader:

• Cut fillers and repeated phrases.
• Split run-on sentences at natural points.
• Add headings at topic changes.
• Check names, numbers and technical terms against the audio.

Turning a video into a blog post and how to summarize a transcript cover the next steps.

Limits and pitfalls

• Speaker changes vanish. Subtitles rarely mark who's speaking, and automatic ones never do, so interviews turn into one voice. Add names by listening, or note them at each change.
• Overlapping or interrupted speech loses its shape when lines are joined.
• Subtitle text may be condensed for reading speed if a person edited it, so it isn't always word for word. Automatic subtitles usually are close to what was said.
• Regular expressions differ slightly between editors; test on a copy first.

Uses for the plain text

Once the words are free of timing, they travel. Common uses:

• Study notes from a lecture, with a timestamp at each topic; lecture video to notes covers turning them into revision material.
• Show notes or a newsletter from an episode; podcast show notes from a transcript has a template.
• A text for a translator or reviewer to read without the clutter of cue numbers.
• Searchable archives, where plain text indexes better than subtitle files.
• Accessibility, as a transcript published beside the video for people who prefer to read.

Each use wants slightly different editing. Notes want headings and bullet points, an article wants polished prose, and an accessible transcript wants every word plus speaker names. Decide the use first and edit for it, rather than polishing the same text three times.