Breakout videos

Search a topic and see which thumbnails beat their own channel’s usual numbers — sorted by how far they beat it, not by views.

Why the breakout factor and not views

A MrBeast video at 30M views tells you nothing about its thumbnail, because the demand exists whatever the packaging looks like. A 143K-subscriber channel whose typical video gets 76K views and whose newest one got 827K is 10.9× — and that thumbnail is worth studying.

So every result is scored against its own channel: that video’s views divided by the median of that channel’s recent uploads. The median, not the average, so one past viral hit can’t redefine what “typical” means. Baseline videos have to be at least two weeks old, because YouTube views are heavily front-loaded and a two-day-old upload in the sample would drag the median down and inflate every multiple on the channel.

Age is accounted for. A video with 100K views in six hours is a bigger breakout than one with 1M after ten days, so a young video’s views are projected forward before it is ranked — views accumulate along a power law, roughly V ∝ t0.65, measured against the live API rather than assumed. Those results show a ~ and the figure they have actually reached so far. Accumulation stops, though, so the adjustment stops with it: past about two days nothing is adjusted, and an old video is never penalised for having had time.

How to use it

  1. Search the topic you’re making a thumbnail for. The box starts from the title in your workspace, so coming here straight from Preview needs no typing.
  2. Read the top of the list. Up to 30 results come back re-sorted by breakout factor rather than by relevance, so packaging that beat its own channel rises whatever the size of that channel.
  3. Send one straight into the AI critic to get the reasoning behind why it works, then borrow the technique rather than the image.

Searching is free: 2 new searches a day without an account, 10 a day signed in. Re-running a search someone has already run today doesn’t count against that.

Two honest limits. Some channels can’t be measured — a channel needs enough recent, non-Short uploads before there is an honest median to compare against, and those results show a dash instead of a made-up number rather than being hidden. And a breakout factor is evidence about packaging, not proof: a video can also break out because of its topic, its timing, or where it got shared.