How to write a video transcript: get one automatically, clean it up, use it to edit

A video transcript is the spoken words of a video written out as text, usually with speaker names and timestamps. Today you rarely type one from scratch. To write a video transcript, get a machine transcript first from YouTube, Word or a transcript-based editor. Then correct it while you listen back, add speakers and timestamps, and break it into paragraphs. If you made the video yourself, the same transcript can also be how you edit it: delete a sentence in the text and the matching clip comes out of the video.

What a video transcript is, and what it is not

A transcript is a document. You read it on its own, or you put it under a video, in show notes or in a blog post. Captions are the same words cut into short timed lines that appear on screen while the video plays. Many tools make both from one speech-recognition pass, which is why people mix the two up.

Before you start, decide how literal the transcript should be:

• Verbatim keeps everything: every “um”, false start and repeated word. Researchers and legal work need this.
• Clean verbatim keeps the meaning and drops the stumbles. Show notes, blog posts and most creator uses need this.

You need to know which one you are writing before you start correcting, because the two lead to different edits.

How to get a video transcript automatically

Which tool to use depends on where the video lives and what you want to do with the text afterwards.

Where the video is	What to use	What you get
Any YouTube video with captions	Show transcript in the video description (YouTube)	The full text. It scrolls with playback, and selecting a line jumps to that point
Your own YouTube upload	YouTube Studio, Subtitles, then Options and Download subtitles (YouTube)	A caption file you can open in a plain text editor
A file on your computer	Transcribe in Word for Microsoft 365, which accepts .wav, .mp4, .m4a and .mp3 (Microsoft)	Text with timestamps and speakers labelled “Speaker 1”, “Speaker 2”
A take you still have to edit	A transcript-based editor such as Descript or slicecut	A transcript connected to the video, so you can cut by editing text

Some limits are worth knowing. YouTube says automatic captions may not be ready right after you upload. It also says they can get words wrong because of mispronunciations, accents, dialects or background noise (YouTube). Microsoft 365 subscribers can transcribe up to 300 minutes of uploaded audio a month in Word (Microsoft).

Upside: a machine transcript takes minutes, not hours, and gets most ordinary speech right.

Downside: it gets the important things wrong: names, brand names, numbers, and words two people say at once. Every automatic transcript needs a correction pass.

How to write a video transcript by hand

Typing a transcript yourself only makes sense for short clips, poor audio a machine cannot handle, or a language your tools do not support. If you have to do it:

1. Play the video at a slower speed, if your player has that setting, and keep the text editor next to it.
2. Type one sentence at a time. Pause, type, rewind a few seconds, and check it.
3. Start a new line with the speaker’s name every time the speaker changes.
4. Add a timestamp at each new topic, or every minute or so, in the form [02:15].
5. Mark what you cannot hear as [inaudible 03:42] instead of guessing.
6. Listen to the whole video once more against the finished text.

How to clean up an automatic transcript

The machine gives you the words. Cleaning up turns them into something people can read. Do it in this order so you only listen through once:

1. Fix names and numbers first. Search the text for each person, product and place you said out loud, and correct every one. Then check every figure against what was said.
2. Add speakers. Swap “Speaker 1” for real names. Check the points where people talk over each other, because that is where tools most often give a line to the wrong person.
3. Choose verbatim or clean. For clean verbatim, take out fillers, false starts and doubled words. Do not tidy up the grammar until it stops sounding like the person who said it.
4. Make paragraphs. Start a new one at each change of thought. A block of text with no breaks is the main thing that makes a transcript hard to read.
5. Keep a few timestamps. One per section is enough for a reader to find their place in the video.
6. Read it once without the video. Anything that only makes sense while you are watching needs a word or two in brackets, such as [holds up the mic].

Use the transcript to edit the video

If the video is your own talking-head take, the transcript is also the quickest map of what to cut. Reading it shows you things scrubbing a timeline hides. You see the sentence you started three times, the tangent in the middle, and the two takes of the same point. Repetition costs a viewer more time than the odd “um”, which is the argument in filler words are not the enemy, repetition is.

In a transcript-based editor, deleting text cuts the video. Descript describes it as editing your recording by editing the transcript: “Delete a word, or every um, and the footage goes with it” (Descript). slicecut works the same way on your phone. Import a take from the gallery and it sends only the audio off for transcription, then returns the words a few seconds later. Tap a word to strike it, or drag across a sentence to strike the whole line. The cut lands right before the next word, and nothing is rendered until export, so every strike can be undone. For a wider look at the options, see Descript alternatives for editing on your phone.

Cutting from the transcript gives you jump cuts, so it helps to know how to make jump cuts that look intentional. Once the words are right, the gaps between them are the next thing to fix: how to remove silence from a video without making it sound rushed. For the full order of passes, from first cut to export, see how to edit a talking-head video.

For a shot where nobody speaks, such as hands at work or a product turning on a table, a transcript gives you nothing to cut by. Use a timeline editor for that.

The short version

• A video transcript is the spoken words written out, usually with speakers and timestamps. Captions are the same words in timed on-screen lines.
• To get one, use Show transcript on YouTube, download the captions in YouTube Studio, use Transcribe in Word, or use a transcript-based editor.
• To write a good one, correct names and numbers, add speakers, choose verbatim or clean, and break the text into paragraphs.
• If it is your own take, edit it from the transcript too: strike the repeats, then tighten the pauses.