How to hook viewers in the first 30 seconds of a YouTube video (what 25 big channels actually do)

What does YouTube measure in the first 30 seconds?

YouTube Studio marks "key moments" on each video's audience retention graph, and its help page defines the first one directly: "Intro tells you what percentage of your audience still watched your video after the first 30 seconds." (Measure key moments for audience retention, YouTube Help, read October 13, 2026.)

• Where to find it: YouTube Studio, Content, the video's Analytics, then the Overview or Engagement tab and the Audience retention report.
• When it appears: "Your video should also be at least 60 seconds long and have at least 100 views."
• What to compare it with: YouTube lets you see "how your video compares to all YouTube videos of similar length" and use typical retention "to compare your 10 latest videos of similar length".

What YouTube does not publish is a target. There is no official "good" percentage at 30 seconds; the third-party figures for that mark run from 50% to 70%, and most come without a named source. I put every public benchmark side by side, with its source, in the retention graph guide. The comparison that means something is your own: this video's Intro against your last 10 of similar length.

How do big channels actually open their videos? (25 channels, 50 videos measured)

The hook guides I read for this page repeat "grab them in the first seconds"; none of them measures what big channels actually do, so I measured it: the two most-viewed recent long-form videos of 25 large channels (MrBeast, Veritasium, Kurzgesagt, MKBHD, Johnny Harris and 20 more), on September 27, 2026. Four findings:

1. The voice starts in the first second in 43 of 50 videos (median 0.08 s; 44 start within 2 seconds, 49 within 5). The one late start is jacksepticeye's "Meme Time" (first word at 20.6 s); jacksepticeye and LEMMiNO are the two channels the detector found opening on a title card.
2. The "cut faster at the start" rule holds in 11 of 24 channels; 6 open slower than the rest of their video and 7 keep the same pace (picture changes in minute one divided by the rest; 0.8 to 1.2 counted as the same pace). Vox and Colin and Samir open 2.5 times faster; LEMMiNO and penguinz0 open at half their usual pace.
3. Cold opens put the title card late. In the four channels where the detector found a cold open followed by a title card, the card lands at 52 s (Veritasium), 89 s (Linus Tech Tips), 112 s (MKBHD) and 150 s (Johnny Harris). Simple History's cold open has music only at the start and the end.
4. Silence is not banned. The longest pause in the voice has a median of 8.6 s, and 35 of 50 videos have at least one pause of 3 seconds or more. A hook needs momentum, not wall-to-wall talking.

How it was measured: voice start = first word in YouTube's captions (about ±0.2 s); picture changes = a frame-difference detector on the video file; cold open = a faster first minute plus a second signal (a title card after the hook, or music only at the start and end). † Gameplay channels: the detector also counts camera movement there, so read their first-minute ratio as indicative. Read the table as a snapshot, not a law: two videos per channel, medium confidence on the opening measures, and it says how these videos open, not how well they retain (I don't have their retention graphs). Faster openers, in order: Colin and Samir 2.50×, Vox 2.50×, Veritasium 2.08×, Johnny Harris 1.61×, DanTDM 1.52×, Marques Brownlee 1.52×, Simple History 1.41×, Linus Tech Tips 1.37×, Ali Abdaal 1.27×, Sidemen 1.25×, Graham Stephan 1.22×.

Download the 50 rows: youtube-openings-2026.csv, free to reuse under CC BY 4.0 with credit to Kevin Tabares, Umbrella Creators. They are the same 50 videos as the loudness study; the editing numbers behind them are summarised in YouTube editing statistics 2026.

Four hook patterns, with a real video for each

These are videos I edited, with what happens in their first 30 seconds as published. I measured the first spoken word with a speech-to-text model run on my own computer (Whisper large-v3-turbo; as a check I put 5 seconds of silence before one clip and the first word moved from 0.0 s to 4.8 s) and looked at one frame per second for what is on screen, so the screen timings are within about a second. View counts are from the YouTube Data API on October 13, 2026. All seven start a voice in the first 0.2 seconds, like 43 of the 50 big-channel videos. None of this says the hook caused the views: topic, title, thumbnail and timing do most of that. It shows what each title promises and the second the cut delivers it.

1. The premise in the first sentence

Say what the title promised, in plain words, before anything else, and show it while you say it. It suits challenge, survival and "what if" videos.

• Puff, 3 Idiots Hunt Down the Sun Man in Minecraft (1,370,403 views, 47:11). First words: "Today me and my friends are hunting down the Sun Man, a giant that grows with the power of the Sun". The Sun Man is on screen from about second 1; the subscribe reminder comes at about 15 s, after the premise, and the hunt starts at about 18 s.
• Crazz, 99 Nights, But Every Day = Pelt Trader Upgrades (820,024 views, 16:38), on a channel of 10.7K subscribers. First words: "So the admin just did something insane." Word-by-word captions from the first frame; by about second 8 a quick montage shows the items, bases and boss fights the upgrades bring.
• Rex & Alexa, ROBLOX FLING THINGS AND PEOPLE WITH ALEXA! (1,366,984 views, 19:51). The game is named at about 2 s, its card is on screen at about 4 s and the first fling happens at about 6 s.

2. The question

Ask the title's question out loud as the first line, then show glimpses of the answer without giving it away. It suits experiments and tests.

• Boffy, How Many Mods Can Minecraft Handle? (1,334,276 views, 22:55), published in December 2024 (earlier work). The first sentence asks how many mods Minecraft can take; by about second 4 the screen shows "1,285 mods loaded" and crash messages, proof that the answer is coming.

3. The cold open: a moment from later, then the story

Open on the strongest moments of the video, labelled or not, and only then start from the beginning. It suits horror, story and long videos whose payoff is far away.

• Swaylemc, It Heard My Voice... (FULL MOVIE) (1,157,567 views, 2:43:30). About 20 seconds of moments from later in the film under a "Coming up..." label, then a "Use headphones" card, and the story starts at about second 27.
• ashlele, DON'T JOIN Dress To Impress AT 3 AM.. (741,166 views, 10:03). About 5 seconds of a game moment, then "hello everyone and welcome back" at about 6 s; the 3 AM premise is said at about 16 s.

4. Headlines first

For news and update videos: read the headlines in the first seconds, then expand each one. Viewers stay for the story they care about.

• Mud, SECRET Skin Update...... (Roblox Rivals News) + New Code (609,848 views, 11:36). "Today on Roblox Rivals News!" is the first line; by about second 11 three stories and a new code have flashed up as on-screen headlines, and the first story starts at about 18 s.

What the seven share: a voice from the first frame, and part of what the title promised, said or shown, before any like reminder or title card.

The first-30-seconds checklist (7 checks)

1. Voice in the first second. 43 of 50 big-channel videos and all seven client videos measured here start talking within it. If yours opens on music, a logo or ambience, test cutting straight to the first line.
2. The first frame matches the title and thumbnail. Puff's Sun Man is on screen from about second 1; Mud's headlines are on screen by second 11. A viewer who clicked for one thing and sees another leaves, and that shows in the Intro metric.
3. No greeting, logo or recap before the promise. The measured cold opens put the title card between 52 and 150 seconds; Mud's like reminder comes after the headlines.
4. Something changes on screen in the first minute. Faster is optional: 11 of 24 big channels cut their first minute faster than the rest, 7 at the same pace, 6 slower. A static opening is the problem.
5. Music under the voice, not on top of it. In the narrated documentaries measured on this site, the music bed sits about 24 to 31 dB below the voice (music volume under a voiceover).
6. A pause is not a failure. 35 of 50 big-channel videos have at least one voice pause of 3 seconds or more. A pause over a frozen image is different: cut that.
7. Check the Intro metric against your own videos. Once the video has 100 views (YouTube's minimum for key moments), compare its Intro with your last 10 videos of similar length, not with a number from a blog.

The longer, minute-by-minute version for the whole video is the retention checklist.

The mistakes that kill hooks

• Too much context before the payoff. "So this is my Roblox account and I've been playing for three years" before anyone knows why they should care. Cut the setup and jump to the moment.
• No clear reason for the video. Within the first lines the viewer should know what they are about to watch and why it matters to them.
• Flat energy in the first seconds. If the opening sounds read from a script, viewers hear it. Record the first line again, or open on a moment instead.
• A hook the video doesn't keep. Promise five ways and deliver one, and viewers leave when they notice; it also costs trust in the next video.
• Logos, watermarks or interface covering the payoff. Branding can wait until after the promise.

How to test a hook without fooling yourself

1. Keep the format fixed. Same series, similar length, similar topic. A hook that works on news says nothing about a tutorial.
2. Run each variant on 3 to 5 videos. One video is an anomaly; several are a pattern.
3. Compare the Intro metric and retention at 60 seconds in Studio against your last 10 videos of similar length, not public view counts. Since August 24, 2026 a public view counts "from the very first frame", while retention and average view duration are still calculated on engaged views (YouTube, August 19, 2026; what changed).
4. Change one thing at a time: the first line, the first image or where the title card goes.
5. Write down what you changed on each video. Three weeks later you will not remember.

Hooks by format: gaming, Roblox, Minecraft horror, documentary

Roblox and gaming. Name the game and the challenge in the first line and show the game at once, as in the Rex & Alexa and Crazz openings above. For news channels, headlines first, as Mud does. More in editing for Roblox YouTubers and Roblox news channels.

Dress To Impress. A game moment, then the greeting, then the premise, as in ashlele's 3 AM video: Dress To Impress channels.

Minecraft horror and full movies. A "coming up" cold open from later in the film, then a short card and the story, as Swaylemc does: Minecraft horror mod channels.

Documentary and faceless. The spy documentary I measured minute by minute cuts its 61-second cold open at about 19 picture changes a minute, then 14.5 in minutes 0–10 and 8.2 at the close (pacing data): the fastest cutting sits where viewers decide whether to stay. See documentary channels and faceless channels.

FAQ

What is a good retention at 30 seconds on YouTube?

YouTube publishes no benchmark. vidIQ reports 38.1% to 67.9% still watching at 30 seconds on six of its own videos, and Subscribr says 50% to 70% without naming a source. Compare each video's Intro metric with your last 10 videos of similar length in YouTube Studio instead.

Should the first minute be faster than the rest?

Not necessarily. Of 24 big channels measured, 11 cut their first minute faster than the rest of the video, 7 keep the same pace and 6 open slower. Something has to change on screen; how fast is a style choice.

Do I need a title card or intro animation?

You do not need one at the start. The four cold-open channels measured place the title card between 52 and 150 seconds in, only 2 of 25 channels open on one, and 43 of 50 videos start the voice in the first second. If you use a card, put it after the promise.

How long should a YouTube hook be?

Long enough to say or show what the title promised. The only official line is YouTube's 30-second Intro metric. In the seven client videos on this page the promise arrives in the first seconds, and the longest opening, a horror film's cold open, hands over to the story at about 27 seconds.

Can an editor fix a weak hook?

The cut can move the payoff forward, remove dead air and put the promise on screen, but the promise has to exist in the footage. Brief the hook before you record: what the title promises and which moment proves it.