Automation Pipeline3 min read

My shorts asked the question and left the answer outside the video

I was auto-generating shorts to promote web apps that give you a result. The question was in the video, but the answer sat behind a link. I switched to capturing the result page URLs directly so the answer appears inside the video, and uploaded the first one.

#shorts#retention#screenshots#youtube#pipeline
Left: a vertical video frame showing only the question and sending the answer to a link. Right: a frame with the result page capture placed inside the video
Same hook. The only difference was where the answer lived.

The question was in the video, the answer was on the site

I build small web apps that give you a result from your input: Korean zodiac animal, star sign, blood type, animal face, and ilju (a Korean birth-date reading). A pipeline turns out a short video for each one. Every video opens with a hook, a line built to make you curious, like which animal your year is or which type you are.

The trouble came right after that. The video made you curious, then never gave you the answer. To see it you had to open the app. From the maker's side it's easy to call that a traffic funnel: people who are curious will click through.

For someone watching a short, though, it's a video that makes a promise and never pays it off. The answer never comes, so swiping away is the natural move.

Does your promo video give viewers the payoff they waited for inside the video, or does it push that payoff behind a link?

Moving the payoff inside the video

The change in this commit is simple: put the answer in the video. The video shows the result screen for the question the hook asked. I call this the in-video payoff.

What would you do? Launch the app and script the inputs until it reaches a result, or look for a place where the result already exists?

Capturing the result URL directly

The new apps have a separate SEO result page for each result: zodiac's sign, ddi's animal, bloodtype's type, animalface's type, ilju's ilju. Since every result has its own URL, I can skip the input screen, open that URL and capture it. Those captures go into the payoff segment of the video.

The new batch lives in its own manifest (apps_hook_web_new.json) and doesn't touch the existing list. The web/shots_new folder where the captures land went into gitignore. The pipeline can rebuild those files any time, so they don't belong in the repo.

The first one: ddi

The first video rendered and uploaded with this structure was the new-design ddi video. This commit covers confirming the pipeline runs end to end and shipping that first video. Nothing more.

Self-check

  • Does the answer to your hook's question actually appear inside the video?
  • Can the result screen you need be opened from a single URL, with no UI scripting?
  • Are capture files you can regenerate piling up in your repo?

The honest part

The idea that putting the answer in the video improves retention is still a hypothesis. I've uploaded one video, and when I wrote this I hadn't checked whether its retention beats the old approach. Showing the full answer in the video could also give people less reason to open the app, and I don't know which way that trade-off goes. What I know for sure is only this: the path from capturing a result page to putting it in the video works.

Related