You cannot find a viral moment, because viral is a fact about the audience, decided after publication. What you can find are the moments that match a pattern you have already seen work, and that is a much more tractable job. The question splits into two: what does a good moment look like for your channel, and how do you locate the ones you have on tape.
Most advice answers the first question with generic rules about hooks. The second question is the one that costs hours.
Read the retention graph on what you already published
If the long video is already up and has views, the analytics page will tell you where people stopped skipping. Spikes in the audience retention curve are the closest thing to real evidence anyone has about which forty seconds of a two-hour video are worth cutting out.
This is the most reliable method here and the most underused. Its limitation is obvious: it only works after the fact, on content that got enough traffic to produce a readable curve. It does nothing for the archive of things nobody watched.
Use live audience reaction as a proxy
For recorded streams, calls, or events, the reaction of the people in the room stands in for the reaction of the people who were not. Chat velocity, laughter, the question everyone asked afterward. It is cheap to capture and reasonably honest.
The failure mode is that it rewards volume. A calm two-minute answer that explains something well produces no spike at all, and those travel well when they get cut.
Watch everything
Nobody recommends this and everybody does it. Playback at 2x, a notes file, timecodes.
It genuinely works, and for a single video it is often the right call. The trouble is that it does not compose. Someone with hundreds of hours of recorded episodes has, in practice, a searchable archive of exactly the parts they happen to remember. Podcast producers with full transcripts online run into this constantly: the material is all there, findable only through search tricks that were never designed for it, so the good moment from episode 180 stays lost.
Let a detection tool rank candidates
Tools in this category ingest the full video and return scored clip suggestions, usually built on some mix of transcript, audio energy, and scene change. They compress a long file into a list you can review in ten minutes.
The tradeoff is that the scoring model has no idea what your audience responds to. You get plausible clips, and you evaluate every one of them yourself. For a first pass on a single video, that is a fair deal. As a way to work through a back catalogue, it is slow, because you review the same volume of candidates for every file.
Search the archive by description
The other approach starts from something you already know. When you can name the thing you are after, the job stops being discovery and becomes lookup: pull up the part where the guest talked about getting fired, find the explanation of the pricing change. Indexing the footage by its content once, up front, turns that into a query that lands on a timestamp with the run-up and the aftermath attached.
Vivu is one implementation of this, and the part that matters for a growing archive is that recordings join the searchable set as they arrive, so nobody has to keep feeding it. What it will not do is tell you which moment deserves to be a short. That judgment stays yours.
When you do not need a tool for this
If you publish one long video a month, the honest answer is to keep a notes file open while you record and mark timecodes as things happen. You will beat any automated system on precision, because you know what you meant. Interview shows benefit the most from this, since the person conducting the interview always knows which answer was the good one.
You also do not need this if your long videos are scripted. The script is already an index.
How to tell which side you are on
Count the number of moments you have used from footage more than three months old. If the answer is zero, discovery is not your constraint and a better recording habit is worth more than any software. If the answer is more than a few, and each one took a search that felt like an archaeological dig, the constraint is that your library has no index, and that stays true no matter how good you get at picking clips.