CueSift取幕
Video subtitle guides

How to Extract Subtitles from a Video

First identify whether the video contains an actual caption track, only spoken audio, or text burned into the image. Then extract the track, transcribe the audio, or use OCR. Choosing the source correctly avoids unnecessary processing and produces a cleaner result.

Published: Reviewed against current CueSift behavior

Short answer

For a public link on a supported platform, paste the URL into CueSift. The service checks for a usable caption track first and can use speech transcription when captions are unavailable and the audio can be processed.

For a local media file, inspect its streams and export the subtitle stream if one exists. If the only visible text is part of the picture, use an OCR workflow instead. CueSift currently does not perform burned-in subtitle OCR.

Check a public video link

Choose the method by subtitle source

The video has a selectable caption track

Extract the caption track

Fastest path. Usually preserves timestamps and language labels.

The video has speech but no usable captions

Use speech transcription

Creates a draft from audio. Accuracy depends on recording quality.

Text is permanently visible in the picture

Use video OCR

Reads image pixels. Requires frame sampling and careful cleanup.

You have a local media file with subtitle streams

Map and export the subtitle stream

A media tool such as FFmpeg can select a subtitle stream without transcribing audio.

Step-by-step workflow for a public video link

01

Confirm access and rights

Open the video in a logged-out or private browser window. If it needs an account, password, payment, or regional access, a public-link extractor may not be able to process it. Only process content you have the right to use.

02

Paste the canonical video URL

Use the individual video page, not a channel, search result, playlist, or generic sharing page. A direct URL makes platform detection and metadata inspection more reliable.

03

Prefer an existing caption track

An existing track usually finishes faster and preserves the creator or platform timing. Review whether the selected language and track type match your needs.

04

Transcribe only when needed

If no usable caption track exists, transcription can generate text from spoken audio. Expect more cleanup when the recording contains music, noise, accents, specialist terms, or overlapping speakers.

05

Review and export

Check names, numbers, punctuation, paragraph breaks, and timestamps. Export TXT for plain text, SRT for broad subtitle compatibility, or VTT for web-video workflows.

How to handle local media files

A local MP4, MKV, WebM, or other container can contain video, audio, subtitle, and data streams. Inspect the stream list before deciding to transcribe the audio. If a subtitle stream already exists, select and export that stream.

FFmpeg documents automatic and manual stream selection, including explicit stream mapping. Subtitle codecs and output containers are not interchangeable, so the target format must support the selected subtitle stream or the stream must be converted.

If the text is burned into the picture, there is no separate subtitle stream to export. An OCR pipeline must sample frames, locate the text, recognize it, merge repeated frames, and reconstruct timing. That is a different workflow from caption extraction.

Frequently asked questions

Can every video subtitle be extracted?

No. A tool needs access to a caption track, an audio source that can be transcribed, or video frames suitable for OCR. Private, deleted, encrypted, or inaccessible sources may not be processable.

Is extracting subtitles the same as transcribing audio?

No. Extraction retrieves an existing text track. Transcription generates new text from speech. Extraction is usually faster, while transcription helps when no text track exists.

What format should I choose?

Choose TXT for plain reading or reuse, SRT for broad subtitle compatibility, and WebVTT when the target web player or workflow requires VTT features.