Paste or drop your file
Paste subtitle text directly into the input panel, or drag in a .srt, .vtt or .sbv file. delsub auto-detects UTF-8 and Windows-1252/ISO-8859-1 encodings, so older subtitle files work too.
100% in your browser
Turn subtitles into clean, readable text
Paste or drop your .srt, .vtt or .sbv file and get continuous, paragraph-formatted text — ready for a blog post, a summary, or a language model. Your subtitle text is never uploaded by the tool: it is processed entirely on your device.
Load a subtitle file
“Roll-up repeats” can remove intentional repetitions. Turn it off when exact wording matters and compare the result with the original. See the example.
Output
The cleaned text appears here and updates as you type. It is a scrollable region.
How it works
Every step below happens locally in your browser: your subtitles are never uploaded by the tool.
Choose what to clean
Toggle timestamps, cue identifiers (SRT numbers and WebVTT IDs), markup and recognized sound cues independently. Rebuild paragraphs to join lines; “Split on speaker” separates recognized speaker prefixes and removes initial dialogue dashes. Your choices are remembered for next time.
Copy or export the result
The output panel updates in real time with a live word count and reading time. Copy it to your clipboard or export it straight to a .txt or .md file, named after the file you loaded.
What the cleanup looks like
A short SRT excerpt with the default options: the sequence numbers, timestamps, italic tags and the [Music] cue go, and the caption lines become sentences.
Input (.srt)
1 00:00:01,000 --> 00:00:03,200 Hello everyone, and welcome back. 2 00:00:03,400 --> 00:00:06,000 <i>Today we're looking at</i> [Music] subtitle formats.
Output
Hello everyone, and welcome back. Today we're looking at subtitle formats.
What people use it for
Subtitles are often the only text version of a video. Cleaned up, they become something you can read, search and reuse.
Articles and show notes
Turn the captions of your own video or podcast into a first draft for a blog post, a newsletter or episode notes.
Study and research
Read and search a lecture or interview at your own pace. Before quoting, compare the passage with the video: cleanup can remove intentional repetitions.
Summaries with AI tools
Give a language model the words without the timing codes: the text is shorter, and the model doesn't treat timestamps and tags as content.
Translation and review
Hand a translator or editor a readable script instead of a timed file, or proofread captions as continuous text.
Guides
How subtitle formats work, and how to get clean text out of them.
- SRT, VTT and SBV: subtitle formats explained How the three common subtitle formats store timing, numbering and styling, and what has to go to leave only the text.
- How to turn YouTube captions into a clean transcript Download a video’s captions or copy its transcript, then remove timestamps, repeats and sound cues.
- Why auto-generated captions repeat every line Roll-up captions explained: where the duplicates come from and how to remove them without losing dialogue.
- Fixing garbled characters in subtitle files Why “é” shows up instead of “é”, how to tell which encoding a file uses and how to convert it to UTF-8.
Questions
Are my subtitles uploaded to a server?
The text-processing code parses and reassembles your subtitles in your browser tab without uploading or storing them on a server. Cloudflare page analytics handles visit and performance data. Advertising is currently disabled. If enabled, Google’s advertising and consent services may handle cookies, identifiers and visit data on this page, and their scripts run in the same page as the tool. Ads also depend on site approval and availability. See the Privacy Policy for details and consent choices.
Which formats are supported?
SubRip (.srt), WebVTT (.vtt) and YouTube SBV (.sbv), plus plain pasted text. The timing option removes valid cue timestamps and YouTube’s inline timestamps. In a file with a WEBVTT header, structural NOTE, STYLE and REGION blocks are removed entirely; the same words inside a cue are kept as dialogue. Markup removal also removes tags such as <v Speaker>.
How does it handle YouTube's duplicated auto-captions?
With “Roll-up repeats” on, identical neighboring lines and lines that extend a previous line are merged. It can also remove genuine repetitions; turn it off for a faithful transcript and review the output against the source.
Can I control exactly what gets removed?
“Cue identifiers” controls SRT indices and text IDs that open a recognized WebVTT cue block. “Sound effects” removes known annotations such as [MUSIC PLAYING], standalone music notes and notes around lyrics, keeping the sung words. Spoken numbers such as “2024”, remarks such as “(see figure 2)” and references such as “[1]” remain. Turn “Split on speaker” off to retain initial dialogue dashes, or “Rebuild paragraphs” off to keep caption lines. WebVTT headers and structural metadata blocks are always removed from recognized VTT files. The formats guide shows the exact behavior and limits.
Why is my text one long paragraph?
Paragraphs are rebuilt from sentence punctuation: a new one starts after a line that ends with a period, question mark or similar. If automatic captions have no punctuation, there is nowhere to break. Turn off "Rebuild paragraphs" to keep one caption per line instead; the YouTube guide explains the options.
Does delsub transcribe or translate videos?
No. It works on subtitles that already exist — a file you downloaded or text you copied — and keeps them in their original language. It doesn't fix recognition errors either: if automatic captions misheard a word, the clean text contains the same word.
Is there a file size limit?
Files up to 10 MB are accepted, which is far more than the subtitles of a feature film (usually around 100 KB). Larger files, and files that aren't text, are rejected before they are read: they are almost always something else, such as the video next to the subtitles.
What's the difference between the .txt and .md downloads?
The text is the same; only the file extension changes. Choose .md if you keep notes in a Markdown-based app, and .txt for anything else. Both are saved as UTF-8 and named after the file you loaded.