Extract Subtitles from Video Free — SRT Online, No Upload

You drop a lecture into a player, pause on a sentence you want to quote, and the letters are already there — timed, spelled, sitting on the picture. You do not need another transcript pass. You need the file those letters came from: an SRT you can search, translate, or mux onto a clean export.

That is a different job from “make captions.” To extract subtitles from video you first have to name the kind of text you are looking at. Soft tracks copy. Burned-in letters do not. Speech with no track at all is a third path, and it is not an extractor.

Most roundups collapse those three into one “online tool” and promise speech-to-text when the container is empty. This page does not. Five methods, one honest split, and a 30-second picker at the end.

3-second answer: If the player can toggle captions, open CutFast subtitle extraction, drop the MP4/MKV/WebM, export SRT or VTT. The file stays on your machine. If there is no track, stop — use auto subtitles or transcribe instead of pretending the extractor invented one.

Table of Contents

• Why extract subtitles from video instead of retyping
• Soft tracks versus burned-in text
• Method 1: CutFast in the browser
• Method 2: FFmpeg when you need every track
• Methods 3 to 5: VLC, Subtitle Edit, VideoSubFinder
• How to choose in 30 seconds

Why extract subtitles from video instead of retyping

Captions are no longer a nice-to-have overlay. Wistia’s 2026 State of Video Report — a survey of about 900 marketers and more than 13 million hosted videos — found that 90% of teams are already taking steps to make videos accessible, and captions are the most common starting point. The same report notes that teams using AI in their workflow are 82% more likely to ship subtitles in multiple languages. More tracks in more languages means more files where the words already exist, timed, inside the container.

You extract because:

• A language learner wants the line, not a blurry screenshot.
• A translator needs the original cue list before they rewrite it.
• An editor is cutting a 40-second quote and does not want to re-time every word.
• You are about to strip the track from a download and still want the text as a sidecar.

If the player can hide the letters, you probably have a file already. Extraction copies that file. Transcription guesses a new one.

Practical rule: Extract when a track exists. Transcribe when it does not. Mixing those two verbs is how “subtitle extractor” pages start lying.

YouTube’s own caption settings Help page is explicit that CC is a player control. A YouTube watch page with CC on is not proof the MP4 you downloaded contains a soft track. Many downloaders bake the selected language into the picture. Pause a frame with every player overlay off before you pick a method.

Soft tracks versus burned-in text

Three labels, used for the rest of this comparison:

Type	What it actually is	Fast test
Player / CC	Overlay from YouTube, the browser, or an app	CC off → letters gone. Downloaded file may have nothing.
Soft subtitle track	A separate stream in MP4 / MKV / WebM (SRT, VTT, ASS, and similar)	Player menu: Subtitle → None. File size barely moves if you drop the track.
Burned-in (hardsub)	Pixels. The letters are in the image	No toggle. Zooming a paused frame still shows the glyphs.

WebVTT is the web-native timed-text format; SRT is the editor-native cousin. Either can live as a sidecar or as a muxed track. Matroska’s subtitle documentation is the map for MKV: text tracks (SRT, ASS) sit next to image tracks (PGS, VobSub). An extractor that only copies text will fail on PGS even though VLC still “has subtitles.”

Burning text into frames is a one-way door. If you still have a separate track, keep it that way until a platform forces you not to. The inverse job is in our online subtitle burner guide.

Practical rule: If you cannot turn the letters off, you do not have an extraction problem. You have an OCR problem, or you are restoring pixels.

Image-based tracks (PGS on Blu-ray rips, VobSub on older DVDs) look like softsubs in a player and still refuse to become SRT in a text extractor. That is Method 4, not Method 1.

Method 1: CutFast in the browser

CutFast’s extract-subtitles tool copies the first exportable text track from a file you drop in the tab. Processing stays on your device. It does not upload the video. It does not fall back to speech-to-text when the container is empty. If that last sentence contradicts an older version of this page, the older version was wrong.

When this is the right method

• The player can disable captions, and you want SRT or VTT in under a minute.
• You do not want to install FFmpeg for a single file.
• Privacy matters: lecture, client review, or anything that should not hit a random “online converter.”

When it is the wrong method

• There is no subtitle stream. The honest next clicks are auto subtitles, video caption generator, or transcribe.
• You need track two of a five-language MKV. Today the tool exports the first track; language pickers are not on this page.
• The “subtitles” are PGS/VobSub images, or they are burned into every frame.

How to run it

1. Open cutfa.st/features/extract-subtitles.
2. Drop an MP4, MKV, or WebM that actually contains a text track.
3. Choose SRT (editors, most subtitle tools) or VTT (web players). The WebVTT spec is why VTT is the better sidecar for HTML5.
4. Extract. Download the file. Confirm the first and last cues against the video.
5. If you need a different container later, run the subtitle converter. If a platform strips sidecars, mux or burn — that is a publish step, not extraction.

Local-first means the extractor can fail with “no exportable track” and still be correct. An empty result is a diagnosis, not a bug.

Practical rule: A clean “no subtitle track” error is more useful than a hallucinated transcript with the wrong names.

Method 2: FFmpeg when you need every track

FFmpeg is the batch and multi-track path. Use it when CutFast’s first-track rule is not enough, or when you are walking a folder of MKVs. The FFmpeg documentation is the source of truth for mapping; the commands below are the ones you actually type.

List streams (look for Subtitle:):

ffmpeg -i lecture.mkv

Copy the first subtitle stream to SRT:

ffmpeg -i lecture.mkv -map 0:s:0 subs.srt

Grab a specific language if the dump showed eng as stream 0:s:2:

ffmpeg -i lecture.mkv -map 0:s:2 -c:s srt english.srt

Image-based PGS will not become a readable SRT with -c:s srt. FFmpeg will copy or fail; it will not OCR. That is still Method 4.

CLI walkthrough of listing streams and mapping a subtitle track. Use this when the MKV has more than one language and the browser tool’s first-track export is the wrong one.

Who this is for: people who already have FFmpeg, or who will run the same command a hundred times. Who it is not for: a one-off MP4 on a phone.

After you have the SRT, the words can leave the editor. If the file is a long talk and you need notes rather than a caption file, BibiGPT is the sister product that turns timed text into a searchable summary — extraction is the input, not the end of the job.

Methods 3 to 5: VLC, Subtitle Edit, VideoSubFinder

These three cover detection, image-track OCR, and hardcoded OCR. None of them replace Method 1 for a clean text track.

Method 3: VLC as a detector

VLC is already on most desktops. It is excellent at answering “is there a track?” and weak at being your extractor of record.

1. Open the file. Subtitle → Sub Track. If you can pick a language or “Disable,” you have a soft track (text or image).
2. If the menu is empty and the letters stay on the picture, you have hardsubs.
3. Do not fight VLC’s convert wizard for a one-off SRT. Once you know the type, go back to CutFast or FFmpeg for text tracks.

VLC will play PGS. That does not mean you extracted SRT. You confirmed a stream exists.

Method 4: Subtitle Edit for PGS and VobSub

Subtitle Edit is the desktop tool for image-based tracks. It OCR’s PGS/VobSub into text, with a human pass. It does not OCR letters that were burned into the video frames. If you cannot disable the text in VLC, skip this method.

Use it when:

• A Blu-ray rip shows subtitles in the player but CutFast reports no exportable track.
• You need to correct OCR before you publish.

Method 5: VideoSubFinder for hardcoded captions

VideoSubFinder is the open tool people actually use for burned-in captions: it finds subtitle-shaped regions across frames, then you OCR the crops. Accuracy follows contrast, font, and motion. A talking-head with a flat bar at the bottom works. Karaoke over a music video fights you.

This is slow on purpose. If you only need one quote, type it. If you need the whole film, budget an evening and a proofread.

If zooming a paused frame still shows the letters, you are in Method 5. Do not shop for an “extract” button on a browser tab.

Practical rule: OCR a burned-in show as a last resort. Recaptioning a trim is often cheaper than restoring a two-hour hardsub.

When the words are finally in SRT and you need them on a platform that strips sidecars, that is adding, not extracting: how to add subtitles online.

How to choose in 30 seconds

Your situation	Method	Install?	Honest limit
Player can disable captions; one file; want SRT/VTT	1 CutFast	No	First text track only; no STT fallback
Several languages in one MKV, or a folder of files	2 FFmpeg	Yes	Will not OCR PGS or pixels
You only need to know if a track exists	3 VLC	Maybe already	Not your export tool
Player has subs; text extractors say none (PGS/VobSub)	4 Subtitle Edit	Yes	Not for burned-in frames
Letters stay after every overlay is off	5 VideoSubFinder	Yes	OCR, then proofread

Decision filter: Spend ten seconds on the pause-frame test before you open any of the five. The wrong method wastes the next twenty minutes.

Five-step finish:

1. Pause with CC off. Letters in the image → Method 5 (or recaption a trim).
2. Player subtitle menu has a track → try Method 1.
3. Method 1 says no exportable track, but VLC still lists subtitles → Method 4 (image track) or Method 2 (wrong stream index).
4. No track at all, but you need words from speech → leave this page. Use transcribe / auto-subtitles.
5. You have SRT. Keep it as a sidecar until a platform forces a burn.

Open the extractor only after the test:

• 📄 Extract a soft track: https://cutfa.st/features/extract-subtitles
• 🎙 No track — generate captions: https://cutfa.st/features/auto-subtitles
• 🔥 Burn only when a platform strips sidecars: https://cutfa.st/features/subtitle-burner
• ✨ More tools: https://cutfa.st/features

FAQ

Q. Why did extraction say there is no subtitle track?

The file has no text subtitle stream, or the stream is image-based (PGS/VobSub) and cannot be converted to SRT here. Confirm in VLC. Then pick Method 4, Method 5, or a transcribe tool — not a second click on extract.

Q. Can CutFast choose which language track to export?

Not today. It exports the first subtitle track. For a specific language, use Method 2 and -map 0:s:N from the stream dump.

Q. Does this page still claim speech-to-text on the extractor?

No. That claim was incorrect. Speech-to-text lives on auto-subtitles, video-caption-generator, and transcribe. The extractor copies a track or it errors.

Q. SRT or VTT?

SRT for editors and most desktop subtitle tools. VTT for web players; see the W3C WebVTT format. Convert later if you guessed wrong.

Q. Timing looks a beat off after extract.

A copied soft track is usually frame-accurate. If you then converted formats, round-trip through the subtitle converter and check the first cue. OCR methods (4 and 5) drift; that is the method, not the player.

Q. What do I do after I have the SRT?

Keep it as a sidecar, mux it, or add it as a new overlay. If you need notes from a long lecture, send the same words through BibiGPT instead of staring at a 2,000-line caption file.

CutFast Team