# How to Format a Full YouTube Transcript for an E-Reader

To read a full YouTube transcript on an e-reader without a wall of text, make a separate reading edition from checked captions. Join short caption fragments into complete thoughts. Add paragraphs, clear headings, a table of contents, and a few source-time links. Keep every substantive passage, including examples and limits. Save the original timed captions separately. Then open the exported document on your reader and check its pages.

The main work is manual editing and page inspection. A caption tool can supply timed text, but you must check its meaning and layout. You do not need to turn the lesson into a summary. You need a readable version of the full text, with a way back to its source.

One public reader described transcripts “without any headings or ways to navigate the text.” That is the problem this method addresses. This is one observed reader problem. It does not establish demand volume or a tested product result. [Public reader discussion, observed 25 October 2026; one reader problem](https://www.reddit.com/r/RemarkableTablet/comments/1t710nf/would_you_read_full_youtube_transcripts_on_your/)

## 1. Choose the source and define the document

Choose one video and one available caption language. A caption cue is a short text passage with a start time and an end time. A caption track is the set of cues for one language. Your reading edition is a separate document made from that track.

Record the source URL, video title, selected language, and the date you collected the captions. Confirm that you can use the content for your intended personal document. Do not publish another creator’s transcript without permission.

Decide that the edition will retain the full spoken content available in the checked track. Paragraphs and headings can change how the text appears. They must not remove explanations, examples, warnings, or conditions. A condition is a limit that changes when a statement applies.

Keep these three records separate:

1. The original timed caption file.
2. The reading edition that you edit.
3. An edit log that records changes and unresolved problems.

An edit log is a record of what you changed and why. It lets you check an edit against the original words later.

If you already have timed captions, you can start with those. If you still need to collect them, use the [caption extraction guide](/en/articles/youtube-caption-extraction/) for that step. This guide covers the reading document and its pages.

## 2. Collect timed captions and check the result

The [YouTube Transcript Scraper, public contract checked 25 October 2026; no runtime test](https://apify.com/chronometrica/youtube-transcript-scraper) reads existing public caption tracks. An Actor is a hosted program that collects data. This Actor accepts direct YouTube video URLs or video IDs. An ID is the short code that identifies a video.

It does not accept a channel, playlist, or search page as a source list. It does not turn audio into new text or download audio or video. It accepts 1–1,000 inputs and processes a repeated video ID once. For this method, start with one video.

Use this small input:

```json
{
  "includeFailedInputDetails": true,
  "outputFormats": ["json"],
  "preferredLanguages": [],
  "startUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"]
}
```

2. Collect timed captions and check the result

Replace the sample URL in `startUrls` with your intended public video URL. The supplied URL is an input example. It is not evidence of a useful transcript or a completed run for this guide.

JSON is a structured text format. Requesting `json` preserves timed segments. A segment is one returned caption cue. Each segment contains `text`, `startSeconds`, `endSeconds`, and `durationSeconds`. Plain text alone does not establish cue end times.

The empty `preferredLanguages` list asks for the best available caption track. It does not promise English. Check `transcriptLanguageCode` after collection. If you set language preferences, the Actor tries them in order. If none is available, it reports a language-unavailable failure. It does not silently substitute another language.

Keep `includeFailedInputDetails` set to `true`. Successful unique videos produce dataset rows. A dataset row is one video result, not one caption or one reading section. Failed inputs appear in the run record called OUTPUT. Do not treat an empty export as a completed transcript.

Check the video identity, language, `status`, and `statusReason`. Confirm that the result contains usable `segments`. Save the original result before editing.

A field is a named value in the result. Some fields can contain `null`, which means that the value is missing or unknown. Preserve that difference. For example, an unknown `videoDurationSeconds` is not a video duration of zero seconds. An unknown `durationSeconds` is not a cue that lasts zero seconds.

The public contract documents this collection method. No paid Actor run was performed for this article. Collection, export, and device checks below are actions for you to perform.

## 3. Join fragments without shortening the lesson

Work on a copy of the caption text. Keep the original segments and their times unchanged.

Read adjacent cues together. A sentence can continue across several cues. Join those fragments before you decide where a paragraph should end. Use a new paragraph when the speaker starts a new thought, example, instruction, or qualification.

Keep the source order. Moving a warning away from the instruction it limits can change the meaning. Keep a definition close to the passage that uses it. Retain a repeated explanation when the speaker uses it to add context or clarify a point.

Follow this editing sequence:

1. Join fragments that form one sentence.
2. Add punctuation only when the meaning is clear.
3. Group related sentences into a paragraph.
4. Compare the paragraph with the source audio.
5. Record any text change in the edit log.

Remove repetition only when you confirm that it is a mechanical caption duplicate. Similar wording does not prove a duplicate. A speaker can repeat a phrase on purpose. If you cannot confirm the cause, keep the passage and mark it for review.

Do not silently replace an uncertain word with a likely word. Keep the uncertainty visible until you can check it. The same rule applies to numbers, names, and technical terms.

A full reading edition keeps the explanation. If you decide to shorten the lesson, treat that as a separate document. The [checked summary guide](/en/articles/youtube-grounded-summary/) covers that different task. Do not label an edition as full text after removing substantive passages.

## 4. Add headings and a source-time map

Create headings at real topic changes. Use short labels that help the reader choose a section. Compare each label with the audio and the screen. A caption phrase can suggest a heading, but the surrounding passage must support it.

State near the start of the document that you added the headings. This keeps an editor’s label separate from the speaker’s words. Do not place an added heading inside quotation marks as though the speaker said it.

Add a table of contents that lists the reading sections. Use linked entries if your chosen document format supports them. Check the entries after export. Keep heading names consistent between the contents and the body.

A source-time map connects each reading section to its original cues. Keep this map even if you show only a short source-time label in the reading edition. It preserves the evidence behind the edited layout.

Save these fields:

| Record | Fields to keep |
| --- | --- |
| Source identity | Video title, `videoId`, `videoUrl`, collection date, and `scrapedAt` |
| Caption track | `transcriptLanguageCode` and `isAutoGenerated` |
| Reading section | Section number, editor heading, and paragraph references |
| Cue mapping | First and last cue references, original start time, and original end time |
| Source access | Checked source-time link, or a written time if a link is unavailable |
| Review state | Checked passages, unresolved passages, and remaining source checks |
| Edit log | Cue reference, original text, edited text, reason, and check result |
| Device review | Export format, review date, page problems, and fixes |

A cue reference is a stable way to identify an original segment, such as its position in your saved file. A paragraph reference identifies a paragraph in the reading edition. Keep both so a later page change does not erase the connection.

Place source-time links at useful section boundaries. You do not need a timestamp before every sentence. Keep the detailed cue mapping in the saved map instead.

Open each link and check where it leads. A source link is useful only when it returns you to the right context. If the reader cannot open it, the visible video title and written time still give you a reference to use elsewhere.

Do not guess a missing boundary. Use returned cue times where available. Preserve gaps and overlaps rather than forcing the cues into neat, continuous blocks. The end of the last caption is not necessarily the end of the video.

## 5. Add checked context where the words need it

Some passages depend on a speaker change, a meaningful sound, or an action on screen. A standalone reading document can need that information.

W3C distinguishes a basic transcript, which includes speech and needed non-speech audio, from a descriptive transcript, which also adds needed visual information. A caption export can therefore need checked additions before it works as a reading document. [W3C WAI: Transcripts, checked 25 October 2026; transcript guidance](https://www.w3.org/WAI/media/av/transcripts/)

Review passages that refer to something the reader cannot see. Words such as “here” or “this” can point to an important screen action. Watch that source passage before you add a note.

Use this process:

1. Identify what the reader needs to understand.
2. Check that information in the video or audio.
3. Add a brief note that clearly says it is an editor’s note.
4. Attach the note to its source range in the map.
5. Mark anything you cannot confirm as unresolved.

Keep added notes separate from spoken text. Do not invent a diagram, speaker identity, sound, or action from caption words alone. If an essential visual detail remains unknown, state that limit. Clear page layout cannot repair missing source information.

This method does not include automatic image extraction. It also does not establish that the Actor checks visual content.

## 6. Build and inspect the reader document

Use a simple document format or PDF that your reader supports. PDF is a document format that preserves page layout. Choose the format from your reader’s documentation. Use the documented import route for that device or account.

Keep the layout simple while you make the first export. Use clear heading levels, readable paragraphs, and a small amount of source information at section starts. Avoid forcing every short section onto a new page before you see how it reads.

Open the exported file on the target reader. A computer preview alone does not complete this check. Inspect:

1. Text size: Can you read the body and editor notes comfortably?
2. Margins: Do words stay within the visible page area?
3. Paragraphs: Do breaks help you follow complete thoughts?
4. Page breaks: Do headings remain with the text they introduce?
5. Navigation: Can you use the contents to reach the intended section?
6. Source references: Are times visible, and do supported links open correctly?
7. Text display: Do punctuation and special characters remain correct?

Check the beginning, a middle section, and the end. Also inspect pages with long paragraphs, tables, or added notes. If a heading appears alone at the bottom of a page, adjust the layout and export again. If a table is too wide, simplify its presentation without deleting its information.

No e-reader import route, device-sync connection, or device layout was tested here. These are proposed checks for your own file and reader.

## 7. Original fiction: a 12-minute reading edition

This worked example is original fiction. The lesson, section choices, and times are invented. It is an illustrative layout, not a real extraction or device test. No complete fictional transcript is supplied, so the table does not prove word-level completeness.

Assume a fictional lesson has three cue groups: 00:00–03:20, 03:20–08:10, and 08:10–12:00. The editor makes three reading sections. The second heading is “Choose an ID.” That is an editor label. The full explanation remains in the body.

| Reading section | Editor heading | Source range | Start seconds | End seconds | Range length |
| --- | --- | --- | --- | --- | --- |
| 1 | Prepare the lesson | 00:00–03:20 | 0 | 200 | 200 seconds |
| 2 | Choose an ID | 03:20–08:10 | 200 | 490 | 290 seconds |
| 3 | Check the result | 08:10–12:00 | 490 | 720 | 230 seconds |

Convert minutes to seconds by multiplying the minutes by 60, then adding the remaining seconds:

- 03:20 = 3 × 60 + 20 = 200 seconds.
- 08:10 = 8 × 60 + 10 = 490 seconds.
- 12:00 = 12 × 60 = 720 seconds.

Subtract each start from its end:

- Section 1: 200 − 0 = 200 seconds, or 3 minutes 20 seconds.
- Section 2: 490 − 200 = 290 seconds, or 4 minutes 50 seconds.
- Section 3: 720 − 490 = 230 seconds, or 3 minutes 50 seconds.

The total is 200 + 290 + 230 = 720 seconds. That equals 12 minutes. This shows that the invented section ranges span the fictional lesson without a gap. It does not prove that every word was retained. Range length also does not measure reading time.

The contents entry “Choose an ID” leads to reading section 2. Its source-time reference points back to 03:20. There is no real video URL for this fiction, so there is no tested playback link.

In a real edition, you would also list the original cues used in section 2. If a cue crossed the proposed boundary, you would keep the complete thought and review the boundary. You would not cut its words merely to match 03:20.

The fictional 720-second lesson length is an example value. It is not a returned `videoDurationSeconds` value. No collection occurred, so `scrapedAt` remains unknown.

## 8. Check completeness and handle failures

Review the full source-to-section map. Each original cue must remain accounted for. A cue can map to a reading paragraph, remain unresolved, or have a logged change such as a confirmed mechanical duplicate removal.

Then compare the beginning, a middle passage, and the end with the audio and screen. These spot checks can reveal errors, but they do not prove full accuracy. Review every uncertain or changed passage as well. Language metadata and caption coverage do not establish correct words.

Use these decisions when a check fails:

| Problem | Next action |
| --- | --- |
| No usable captions or a failed input | Record the failure. Stop the edition until you have a usable source. |
| Requested language unavailable | Choose another available track only if it meets your need. Record the change. |
| Unclear words | Keep the passage unresolved until you can check it. |
| Missing visual context | Inspect the video and add a checked note, or retain the stated limit. |
| Unmapped substantive passage | Restore or map it before calling the edition complete. |
| Unreadable pages or broken navigation | Fix the export and inspect it again on the reader. |

Keep public source evidence separate from records available only in your account. The public Actor contract describes supported inputs and outputs. Your run’s OUTPUT and saved files document what happened in your own collection. Do not assume they are public evidence.

The named public preset was present when checked on 25 October 2026. That establishes public page presence only. Its saved account input and execution were not checked. It is a documented starting point for collection, not evidence of a finished reader edition.

## Make and check one reading edition

Start with one known video. Keep its original cues, edit a separate reading copy, and save the source-time map and edit log. Inspect the exported pages on your reader before preparing more documents.

If you need timed captions, open the [YouTube Transcript Scraper](https://apify.com/chronometrica/youtube-transcript-scraper). Replace the sample URL with your source and request JSON. After collection, decide whether the track is usable. If it is, continue with the reading edition. If essential content remains unresolved, keep that limit visible before you rely on the document.

## Sources

- [Chronometrica / Apify: YouTube Transcript Scraper — public contract checked 25 October 2026; no paid run or runtime-health test](https://apify.com/chronometrica/youtube-transcript-scraper)
- [Apify: youtube-transcripts-plain-text — public page and Task ID checked 25 October 2026; saved input and execution not tested](https://apify.com/chronometrica/youtube-transcript-scraper/examples/youtube-transcripts-plain-text)
- [W3C WAI: Transcripts — checked 25 October 2026; guidance on speech, needed sounds, and visual information](https://www.w3.org/WAI/media/av/transcripts/)
- [Public reader discussion — published 8 May 2026 and observed 25 October 2026; one reader problem, no demand or performance evidence](https://www.reddit.com/r/RemarkableTablet/comments/1t710nf/would_you_read_full_youtube_transcripts_on_your/)