Key takeaways
- Start from what already went viral, then look for the moments around it, rather than scrubbing blind.
- Score every candidate on one-second readability, payoff inside 30 seconds and fit with the campaign brief.
- A model or transcript search can timestamp a three-hour video in minutes, but you still make the final pick.
- Current and on-brief beats bigger: off-audience clips get rejected no matter the views.
Find viral moments in long videos by starting from what has already spread, then scanning for reaction spikes, scoring each candidate against three tests, and only then cutting. Do not start by scrubbing the timeline. That is how three hours disappear.
This is a repeatable method with a scoring sheet. It works for podcasts, streams, interviews, and sports broadcasts you have the right to clip.
Step 1: Start from what already worked
The strongest shortcut is to look before you watch. If someone has already posted a moment from this person or show and it spread, that tells you what the audience wants more of.
Our own process for celebrity and creator content begins on X. We search for the person's name with a likes threshold and read the top replies. That gives a list of moments people already shared. We pull the media from those posts and rebuild them for the format we need.
Our records also state a plain rule: check what is actually winning first, then copy the shape. Query the top-viewed recent posts, look at them, and then build.
For your own workflow:
- Search the person or show name on the platform you post to.
- Sort by top, filter to the last few weeks.
- Write down five moments that spread, and what they have in common.
- Look for similar moments in the long video you have.
Now you are hunting for a type of moment, not a random one.
Step 2: Scan for signals
Run through the video at 2x speed while you watch for signals. Use a text note for timestamps.
| Signal | Why it matters |
|---|---|
| A loud or surprised reaction | Viewers feel it before they know why |
| Silence after a line | A joke landed or a claim shocked |
| A number or a claim | Gives people something to argue with |
| A mistake or a recovery | Humans are drawn to both |
| A quotable one-liner | Easy to caption and share |
| A topic change that is abrupt | Often signals a story worth cutting |
| Chat or comments flooding (for streams) | Many different people reacting is a stronger sign than one person spamming |
The more signals stack in one spot, the better the candidate.
Step 3: Use a transcript or a model
Long audio is where tools earn their keep.
- Transcript search. Search the transcript for the signals: numbers, strong words, laughter tags, names. It turns scanning into searching.
- A model with audio input. We have used Gemini with audio input to timestamp long podcasts. You give it the audio and ask for moments with timestamps and a reason. It narrows hours into a list of candidates.
- Opus-style clip tools. Tools that auto-suggest clips exist. See our guide on how to clip with Opus Clip. Treat their picks as candidates, not decisions.
Two warnings. First, models and auto-captions get names and numbers wrong, so check them against the audio. Second, a tool cannot tell whether a moment fits a campaign. That is your job.
Step 4: Score each candidate
Before you cut, score. It takes 30 seconds per candidate and saves a lot of wasted edits.
| Test | Pass if |
|---|---|
| One-second read | A stranger understands the clip in one second at phone size |
| Payoff speed | The payoff lands inside 30 seconds |
| Hook available | There is a strong first second to start on |
| On brief | It fits the campaign's niche, season and rules |
| Fresh | The moment is current, not a throwback, unless the brief wants one |
| Safe | No gore, injury, anything distasteful, or content the brief bans |
A candidate needs to pass at least the first four. If two fail, skip it.
Worked example
You have a three-hour podcast. A transcript search plus a model returns eight candidates. You score each on the table. Three fail the one-second read because they need background. Two are off-brief. Three pass. You cut those three, and the best one becomes post one. Total time: about 45 minutes for three clips, rather than a blind three-hour scrub.
Step 5: Pick current over bigger
A rule we keep running into: current beats bigger. On a current-season pitch, throwback clips had the biggest view counts and still got rejected. For sports, that means the current season. For culture and news, it means this week.
There is a second filter: audience fit. In our curation, a clip that is off-audience for the brand is rejected no matter the views. If a campaign wants finance content, a random funny moment is not on-brief even if it has more views. A finance campaign takes finance content only, a poker campaign poker only. Read the brief first, then hunt.
Step 6: Test the hook before you cut
Place your cursor where the clip would start. Ask:
- Does the first second give a reason to stay?
- If not, can I start later, on the reaction?
- Or is a text hook enough?
If you cannot find a hook, the clip is not ready. Our post on captions and hooks for clips gives patterns. A short length helps; see best clip length for TikTok.
What kills a pick
- Setup-heavy moments. If it needs ten seconds of background, the viewer is gone.
- Inside jokes with no payoff for strangers.
- Old moments presented as new. Check dates.
- Content you do not have rights to. Use only what the campaign provides or permits and check the creator's official channels.
- Clipping the whole thing. A clip should be one idea.
A fast routine for any long video
- 5 minutes: search what already spread for this person or show.
- 15 minutes: scan at 2x or search the transcript, noting timestamps.
- 5 minutes: score candidates on the table.
- 10 minutes: cut the top one or two.
- Post, then check the first-hour result and learn which signals paid off.
After 10 clips, compare which signals predicted the best views. Over time you build your own sense for it.
Where to apply this
The same method works for podcast clipping, clipping streamers and live streams and turning a YouTube video into Shorts and Reels. For which videos are worth clipping at all, see best types of videos to clip. Campaign rates are set per campaign and vary, and the live clipping rates page shows what the market advertises. Nothing here promises views or earnings.
Frequently asked questions
How do you find viral moments in long videos?
Start with what has already spread: search for the person's top posts and read the top replies. Then scan the long video for reaction spikes, quotable lines, numbers and mistakes. Score each candidate for one-second readability, payoff in 30 seconds and brief fit, then cut the best.
Can AI find the best clips in a video?
AI can help. We have used a model with audio input to return timestamps from long podcasts, which saves hours of scrubbing. It narrows the list. You still check each candidate by watching it at phone size and making the final call.
What makes a moment go viral?
A moment spreads when a stranger understands it in one second, the payoff lands fast, and it gives people something to react to or argue about. Current, on-audience content travels. A famous moment from years ago usually does not, unless the angle is fresh.
How long does it take to find a clip in a 3-hour video?
With a plan, about 15 to 30 minutes: 5 minutes to check what already spread, 10 to 20 minutes to scan at 2x or search a transcript, and a few minutes to score. Blind scrubbing can take the full three hours.
How do I know if a moment is on-brief for a campaign?
Read the brief first. Campaigns often accept only a niche, a season or a format. Off-brief clips can be rejected whatever their views. Check the allowed content, required captions and any banned topics before you pick.
Get paid for the views you already get.
New campaigns drop in the FindClout Discord first. Join it, then join FindClout free and apply once. We place your page on campaigns that fit it, and you get paid per verified view.
Get Aura Farm in your inbox
One email when a new piece drops: clipping, distribution and where attention is going. No spam.