Export Selected YouTube Captions to Markdown
For students, researchers and writers: use Vidleaf to save the caption lines you select, with source information and timestamps for your notes.
Free scope: reading existing captions, transcript search and selected Markdown exports are free. Guests have a daily limit on new reads; a registered account has unlimited reading of available captions. New transcription, translation and summaries use separate allowances and require sign-in. See current allowances.
By Vidleaf Docs, the team behind Vidleaf · Published October 7, 2026 · Updated October 9, 2026
How do I export selected YouTube captions to Markdown?
In Vidleaf, open an available caption track, select the complete passage, then choose “Export” → “↓ Selected Markdown”. The file keeps the video title, one source URL, caption-source label and a plain-text time range for each selected row. Open the downloaded file and check that its wording and selection match the reader.
Jump to the downloadable real export and its format, or work through the full selection example below.
A useful research note answers one question and lets you retrace the answer. This guide turns a short section of a YouTube recording tutorial into a note with a source, a time range, an unchanged caption export and a separately labelled paraphrase. You can repeat the same steps for a lecture or technical talk.
You need a public video with an accessible caption track and somewhere to save Markdown or plain text. If you have not found the passage yet, first search a YouTube transcript for an exact quote. Existing captions are enough; translation, transcription and an AI summary are optional. The examples below use English captions and English interface labels.
1. Check the source and caption track
Start with a question: Why would an interviewer record two voices into separate channels? Our example is Best Practices & Tools for Oral History Recording — Tutorial. Keep its title and YouTube URL in your note. If you want to inspect captions on YouTube first, its transcript instructions describe the “Show transcript” option in a video's description.
Source credit: Tulane University Libraries, published August 28, 2025, under CC BY 3.0. We reproduce a short caption excerpt below and label our own commentary separately. No endorsement by the source publisher is implied.
Paste the URL into Vidleaf's transcript tool and choose “Check available captions”, or open the saved English result. Check the language and source label before reading. A human-caption label identifies the track type; it does not certify every word. Changing the interface to English does not translate a video.
A public video is not a guarantee of accessible captions. Private, removed, age-restricted or region-restricted content may be unavailable; platform errors can also interrupt a caption request. If the check confirms no captions, review the offered transcription option, sign-in requirements and current allowance before confirming. A failed request alone does not establish that captions are absent.
2. Search for the wording, then read around it
Open the reader's “Search” tab. Enter "split mode", including straight double quotation marks, and press “Search”. Without quotes, spaces separate keywords; with quotes, the words form one search term. Short terms can match inside longer words. This searches caption text, so an idea expressed with different wording may not appear.
The saved English track returned three matches for the phrase on October 7, 2026. Open the result near 14:28 by clicking its text to see the transcript context. A result's timestamp opens the original video. Under “Search options”, increase the context lines when a match appears to start halfway through an explanation.
If there is no match, try channels or another word from the passage. Search can match a phrase across neighbouring caption rows and ignores case and repeated whitespace. It still searches the saved wording: punctuation, caption errors or a different phrase can prevent a match. Check that you selected the intended language track and display mode: searching original text, translated text and a bilingual view can produce different results.
3. Select the whole thought, not just the matching row
In “Transcript”, read the lines before and after the match. The phrase names a recording mode, but the next sentence explains its purpose. For this example, select the four consecutive rows starting at 14:25, 14:28, 14:32 and 14:34. Together they cover two sentences ending around 14:38.
Use each row's checkbox. Check the selected-line count before exporting; a previous selection on another page can still be included. To start cleanly, use “More” → “Clear selection”, then check these four rows. Do not select the entire transcript when your note needs only this explanation.
Listen from a little before 14:25 on the source video. Establish what “this compact system” refers to, then hear the statement about separate channels and independent volume adjustment. Keep the condition and purpose together. This passage describes a recording setup; it is not evidence that automatic speaker detection always identifies people correctly.
4. Export the selected text with its source
With the original text displayed and four rows selected, open “Export” and choose “↓ Selected Markdown”. Save the file next to your note. The current Quote Markdown export keeps the video title, source URL, caption-source label and a time range for each selected row. It preserves the caption wording rather than rewriting it into a summary.
Download the real selected-caption sample (.md)
This is the unchanged four-row English file downloaded from Vidleaf on October 7, 2026 without signing in. It covers 00:14:25–00:14:38 of the Tulane University Libraries tutorial credited above. It contains a short excerpt, not the full transcript, and no translation or AI summary. The file preserves its original export date.
Current format, checked October 8, 2026: the September 24–27 examples elsewhere used a separate video link after each row. The current file retains a single source URL at the top and the selected rows' time ranges. Changing the export layout does not rewrite the selected caption text.
The timestamps in this export are plain text; the source URL is at the top. Do not expect a separate clickable video link after every quotation. The reader's timestamp links remain useful for checking a passage. SRT preserves milliseconds; the Markdown display rounds times down to whole seconds.
“Unassigned” in an export means no speaker label has been assigned. Do not replace it with a person's name unless you have established that identity. If you switch to a translated or bilingual display, check the exported language and retain any machine-translation label. To keep both languages in your study notes, follow the guide to bilingual notes from a YouTube lecture.
Export limits: only the selected rows are included. A download does not fix inaccurate captions, fill missing words or verify a quotation. If the source track has changed, reload and reselect. A bilingual export needs a saved translation aligned with that same track; if it is unavailable or stale, use the original-only export or review the translation option.
For this run, the four-row export spans 00:14:25–00:14:38. Reopen the downloaded file and check its first and last lines, source and selection count against the reader. Keep this original file when you tidy a copy for your notes.
5. Keep evidence and interpretation separate
A short note can be more useful than a full transcript. The card below is our own example note. Its question, paraphrase and follow-up are written by Vidleaf; the four caption rows and time ranges reproduce the exported passage. The original wording is preserved.
Recording two voices for later editing
Question: Why record the speakers into separate channels?
Source: Best Practices & Tools for Oral History Recording — Tutorial, English human captions, 14:25–14:38.
Caption excerpt, unchanged; times from the Markdown export:
00:14:25–00:14:28
This compact system offers users the flexibility00:14:28–00:14:31
to record each speaker into separate left and right channels in “split mode.”00:14:32–00:14:34
Enabling this feature is particularly handy for adjusting00:14:34–00:14:38
the volume levels between each voice independently in post-production.
Our paraphrase: The tutorial describes keeping the two voices on separate channels so their levels can be adjusted independently later.
Follow-up before using this setup: Check whether the actual recording equipment supports that mode and make a short test recording.
Evidence retained: The downloaded Markdown file with its source link and caption-source label, alongside this note.
Do not put your paraphrase in quotation marks. If you join caption rows, preserve their words and order; mark omissions or corrections so readers can distinguish editing from the source. Check names, numbers, negatives and technical terms against the audio. YouTube itself warns that automatic captions can misrepresent speech, including when accents or background noise interfere.
An AI video summary can help locate another topic, but it is generated text. Keep it in a separate section and verify its claims before incorporating them into your own writing. “Selected Markdown” exports caption selections; “Markdown notes” is a different export associated with the summary. Neither supplies independent evidence beyond the underlying video.
6. Apply the same method to two harder cases
A number needs its scope
Open Building AI Agents at Scale and search million. Read the clarification at 20:16–20:43: the reported core count covers shared platform services, including APIs, pipelines, web applications and agents. A note that assigns all those cores to agents changes the claim. Attribute the number to the speaker and keep its scope; do not present it as an independently measured benchmark.
An explanation may need eleven rows
In Getting Started with Python in Visual Studio Code, search "virtual environment". Read the application A/B example at 10:19–10:45: different package requirements can cause conflicts in a shared environment. The demonstrated selection contains eleven rows. Keep the cause and consequence together, and treat the recorded editor interface as historical. Check current documentation before copying commands or relying on menu locations.
These three examples were checked against the saved English tracks and selected Markdown exports on October 7, 2026. Caption text, timing and match counts can change. Use the passage wording and surrounding context as your guide.
Read the chapters and source passages
- Oral history: separate recording channels. Follow the recording tutorial's chapter notes and the complete split-mode selection example.
- AI agents: check what the scale figure covers. Read the shared-platform clarification before using the million-core figure in a note.
- Python: keep the package-conflict example together. Use the chapter notes to find the full application A/B explanation and its export example.
When the workflow does not behave as expected
| What happened | Next step |
|---|---|
| The video plays, but the caption check fails. | Confirm the URL and try YouTube's transcript view. Keep the distinction between an unavailable track and a failed request. Retry later if the platform is temporarily unavailable. |
| Search returns too much or nothing. | Quote a multiword phrase; if no result appears, try a distinctive shorter term and inspect nearby rows. Check the language track and original/translated display. |
| The export contains unrelated passages. | Clear the previous selection, select only the intended adjacent rows and check the count again. “All TXT” and “All SRT” export the full track. |
| The wording or selection has changed. | Reload the transcript, locate the passage again and review it before exporting. Keep your older file labelled with its date rather than treating both versions as identical. |
| Markdown download is unavailable. | Confirm that rows are selected. Try “Selected TXT” as a fallback and add the source URL yourself. Review any error message before repeating the action. |
Before sharing, check that the source opens, the quote says what you claim, and your own words are clearly labelled. Keep attribution and use only the excerpt your explanation needs. See Vidleaf's content rules and data-handling information.