Skip to content

Clip finder: the best moments in a transcript

Paste a YouTube transcript, or open an SRT or VTT file, and get the stretches most worth clipping, best first, each with the words that put it there. Then cut them in a click, or watch each one on YouTube at its time.

A free tool from Clyper, the AI clipping tool that turns long videos into captioned vertical clips.

Get started

On YouTube: under the video, open the description, choose Show transcript, then select it all and copy. Or paste any transcript with times, or open an SRT or VTT file.

Read as a transcript with times: 54 lines, 3 min 49 s · a made-up example episode

Length

Moments

5 moments worth clipping · best first

  1. 2:50–3:3141 s+15 points

    “Let me tell you about the worst week we ever had. One time we posted 21 clips in seven days and got almost nothing.”

    +3Opens on a bold claim: the worst+3Opens on a story: let me tell you, one time+2Numbers: 21, 1,000+2Superlatives: worst, biggest+2Laughter or applause: [laughter]+1A turn: turns out+1Strong feeling: furious+1Ends on a full stop

    All its words

    Let me tell you about the worst week we ever had. One time we posted 21 clips in seven days and got almost nothing. Not one of them went over 1,000 views. I was furious. [Laughter] But it turns out the problem was the first second of every clip. Every clip started with "so" or "um". Nobody waits for that. We fixed one thing, the opening words, and the next week we were back. That's the biggest lesson of the whole year, honestly. Start on the sentence that matters.

  2. 0:29–1:0940 s+11 points

    “Have you ever posted something you were sure would flop, and it became your biggest video ever?”

    +3Opens on a question+2Numbers: 40, 2 million, million+2Superlatives: biggest, best+2Laughter or applause: [laughter]+1Strong feeling: crazy+1Ends on a full stop

    All its words

    Have you ever posted something you were sure would flop, and it became your biggest video ever? That happened to me in month three. I posted a 40-second clip at midnight, almost as a joke. I woke up and it had 2 million views. Two million. My best before that was maybe 900. [Laughter] And the crazy part is, it wasn't even the best moment of the episode. It was the opening line that did it. That's the whole lesson. Right, right. So what was the line?

  3. 1:45–2:2439 s+9 points

    “Here's the thing most people get wrong about clips. They cut the moment where the story ends, not where it starts.”

    +3Opens on a bold claim: here's the thing, most people+2Numbers: 30, 45, 60+1A turn: actually+1Superlatives: best+1Strong feeling: wild+1Ends on a full stop

    All its words

    Here's the thing most people get wrong about clips. They cut the moment where the story ends, not where it starts. You need the setup. Without the setup, the payoff means nothing. Actually, the best clips are 30 to 45 seconds, setup to payoff. I tested 60-second versions against 30-second versions for a month. The 30-second ones got twice the watch time, every single time. That's wild. It is. And you don't need fancy edits. You need a hook and a payoff.

  4. 1:09–1:4536 s+5 points

    “I said, "Nobody tells you that the first 100 videos are practice."”

    +3Opens on a bold claim: nobody+1Numbers: 100+1Ends on a full stop

    All its words

    I said, "Nobody tells you that the first 100 videos are practice." And people stopped scrolling, because it's true. Yeah. Um, so, and then, like, we kind of kept going with it. We did a few more like that. And, uh, it was fine. It was okay. Then we changed the setup a bit and moved the camera. Mm-hmm. Anyway.

  5. 2:27–2:5023 s0 points

    “Most people watch with the sound off, so your words have to be on screen. I put big captions on everything now, word by word.”

    +3Opens on a bold claim: most people+1Ends on a full stop-4Housekeeping: leave a comment, follow us on, merch

    Few signals in its words: the weakest of these picks.

    All its words

    Most people watch with the sound off, so your words have to be on screen. I put big captions on everything now, word by word. Okay, quick break. If you're enjoying this, leave a comment below and follow us on Instagram. We read every comment, and the merch is in the link below. Alright, we're back.

This counts words. Clyper understands them. Its AI reads the whole conversation, picks the moments that stand on their own, cuts each on whole sentences, frames whoever's talking in 9:16, captions every word and scores every clip's hook, from the YouTube link alone.

How to find the best moments to clip

  1. Copy the transcript. On YouTube, open the video's description and choose Show transcript, then select it all, times included, and copy it. A podcast app's transcript with a speaker and a time over each turn, or an SRT or VTT file, works too.
  2. Paste it in. The clip finder reads it and lists the moments, best first. Choose how long they should be (15–35 s, 20–60 s or 45–90 s) and how many you want.
  3. Check each one. Add the video's link and each moment opens on YouTube at its time. Keep the ones that make sense on their own.
  4. Cut them. Copy them as timestamps, or send them all to the free video clipper, which cuts every one from the video file at once, as it is or vertical.

What it looks for

A clip is decided in its first seconds and kept by what comes after. So the clip finder scores each possible stretch on how it opens (its first 6 seconds) and what's in it, adds the points up, and keeps the best stretches that don't overlap. No single kind of word can carry a moment: each adds a few points at most.

The signals the clip finder counts, and their points
SignalPointsWhat sets it off
Opens on a question+3A question mark, or "why", "how", "have you", "would you" in the first 6 seconds
Opens on a bold claim+3"Nobody", "never", "the truth", "most people", "here's the thing", "the biggest" in the first 6 seconds
Opens on a story+3 (+1 later on)"Let me tell you", "one time", "I remember", "years ago", "turns out"
Laughter or applause+2 to +3[Laughter], [Applause], (laughs), "haha"
Numbers+1 each, up to +2"2 million", "$500", "30%", "twice", "half"
A turn+1 each, up to +2"Actually", "turns out", "instead", "the problem is", "the catch"
Superlatives+1 each, up to +2"best", "worst", "biggest", "craziest", "of all time"
Strong feeling+1 each, up to +2"crazy", "furious", "no way", "unbelievable", "!"
Talks to the viewer+1 or +2"You" and "your" as a twentieth of the words or more (a tenth for +2)
Ends on a full stop+1The last line ends a sentence, or on laughter
Sponsor read-6"Use code", "brought to you by", "today's sponsor", "20% off"
Housekeeping-4"Subscribe", "link in the description", "welcome back", "thanks for watching"
Starts on filler-1A first line of only "yeah", "right", "anyway" or [Laughter]: the moment starts a line later
Starts mid-thought-1A first line starting "and", "but", "because" or "which"
Few words-2Under 1.2 words a second: music, silence or a long pause

Every result shows the signals it found and the words that set them off, so you can see why it's there and overrule it. The opening signals are the ones that matter most, for the reason in short video hooks.

What it finds in the example

The example is a made-up episode of a made-up show, a little under four minutes long, with a sponsor read and a “subscribe” in it. At 20–60 s, these are its three best moments:

The clip finder's three best moments in its example transcript
MomentOpens onPointsWhy
2:50–3:31“Let me tell you about the...”+15opens on a bold claim, opens on a story, numbers
0:29–1:09“Have you ever posted something you...”+11opens on a question, numbers, superlatives
1:45–2:24“Here's the thing most people get...”+9opens on a bold claim, numbers, a turn

The opening minute, with the welcome, the sponsor and the call to subscribe, comes nowhere near the top. That's the part most clips should never include.

What a transcript can't tell you

The clip finder reads words, so it's blind to everything else that makes a clip work. It can't see a face, a reaction or a screen, can't hear a voice rise, and can't tell a joke from a statement. It hears laughter only when the transcript writes it down, and it can't know whether a moment makes sense without the minute before it. That's what the shortlist is for: watch each moment and keep the ones that stand alone.

This is where Clyper is different. Its AI reads the whole conversation the way an editor would, picks the moments that stand on their own, and turns each into a finished clip: cut on whole sentences, framed in 9:16 on whoever's talking, captioned word by word and scored for its hook, up to 30 from one YouTube link. How that works is in how AI clip generators work.

Other ways to find clips

By sound. The video highlight finder listens to a stream or podcast file and marks where it gets loud and lively: laughter, shouting, hype. It's the better pick for a gaming stream, where the best moments aren't always in the words.

With ChatGPT. Paste the same transcript into ChatGPT or Claude with a good prompt and it can suggest moments too, with reasons. The prompts that work, and what to check before you trust its times, are in how to use ChatGPT to find clips.

No transcript? The video to text converter writes one from the video's speech, up to 30 minutes at a time, on your device. For everything else on choosing, read how to find the best moments to clip.

Questions

How do I find the best moments in a video to clip?

Read it rather than watch it. Copy the video's transcript into the clip finder and it marks the stretches of 20 to 60 seconds that open on a question, a bold claim or a story and carry numbers, a turn, laughter or strong feeling, with sponsor reads and "subscribe" pushed to the bottom. Then watch only those moments and keep the ones that stand on their own.

How do I get the transcript of a YouTube video?

On a computer, open the video's description and choose Show transcript, then select the whole transcript with its times and copy it. Any video with captions has one. For your own video, a downloaded SRT or VTT caption file works too.

Is the clip finder free? Does it use AI?

It's free, with no sign-up, and it doesn't use AI: it counts plain signals in the words, the same way every time, and shows you which ones it found. It runs in your browser, so the transcript isn't sent anywhere.

What makes a moment worth clipping?

It stands on its own and hooks in its first seconds: a question, a claim people will argue with, or the start of a story, then a payoff before it ends. Numbers, a twist and a laugh help. A clip that needs the minute before it to make sense, or opens on a sponsor read, doesn't work.

How long should a clip be?

Long enough to hold the setup and the payoff. The clip finder looks for 15–35 s, 20–60 s or 45–90 s; Clyper's clips run 20 to 60 seconds, cut on whole sentences.

Can it find clips in a podcast or a Twitch stream?

Yes, from a transcript with times. A podcast on YouTube has one under its description; for a file without one, the free video to text converter transcribes up to 30 minutes at a time on your device. For a long stream, the video highlight finder finds the liveliest moments by their sound instead.

Does it work in other languages?

Partly. Its word lists are English, so in other languages it only counts question marks, numbers, exclamation marks and laughter or applause cues. Clyper's AI picks moments in the video's own language.

How is this different from Clyper?

The clip finder counts words and gives you times to cut yourself. Clyper reads the whole conversation with AI, picks the moments that stand alone, cuts them on whole sentences, frames whoever's talking in 9:16, captions every word and scores each clip's hook: up to 30 finished clips from a YouTube link.

Sources

Where YouTube's transcripts are, from YouTube's help page, read on November 13, 2026. The signals and their points are this tool's own, set out above.

Skip the shortlist: get the clips

Paste a YouTube link and Clyper's AI reads the whole video, picks its best moments and makes up to 30 vertical clips, cut on whole sentences, captioned and scored for their hook.

Works with
  • YouTube
  • Twitch
  • Google Drive
  • Dropbox

or upload a video

See plans and pricing

Guides

More free tools