Features › A/B testing variations

Generate variations to A/B test

The thumbnail you like is not reliably the thumbnail that gets clicked. The only way to know is to put more than one in front of an audience. Generating a second, third and fourth take costs a credit each, which is cheap enough that guessing stops being worth it.

Try it, 3 free See pricing No card to start. Packs from $4 after that, nothing recurring.

The same scene twice side by side, marked A and B, one graded cold and flat, one warm and high contrast — **Made with ThumbnailHammer.** One variable changed between the two: the grade. Everything else held, so the test result means something.

How it works

Generate your first take

Describe the thumbnail and make it.

Change one thing and run it again

One variable at a time is what makes the result readable: the expression, the headline, the background, the crop. Change three things at once and you learn nothing.

Test them and keep the winner

YouTube's own Test & Compare will rotate thumbnails on a live video and report which one won. Feed the winner back in as a reference for the next video.

Things people ask it for

Same shot, but the headline reads IT BROKE instead of WHY IT BROKE

Same scene, one version with my face and one without

Same idea in three colour schemes: red, cold blue, high-key white

Questions

Does a variation cost extra?

It costs one credit, the same as any other generation. At the 50-credit pack that is 30 cents a variation.

How many should I test?

Two or three per video is plenty. YouTube needs a few thousand impressions before a difference means anything, so more variants mostly slows the answer down.

Where do I run the test?

In YouTube Studio. ThumbnailHammer makes the images; YouTube decides the winner on real traffic.

Also in the builder

Three free, no card.

Make one and see whether it beats what you have now.

Launch the builder