Iterating on an AI Promo Video: What Changed Across 5

A comment on my previous video was blunt and fair: "If you'd shown the result upfront, nobody would have watched — it's terrible." Instead of brushing that off, I went back and reworked the video from scratch.

A Quick Recap of the Project

What I was trying to promote is an automation system I built: whenever I post a Reel on Instagram, it automatically converts the video into a Turkish and English blog post on my website. The tool works well — the first promo video just didn't do it justice.

Round One: Starting with a Vague Note

My first revision prompt was something like: "Better, but the music is too generic and the transitions don't make you go 'wow.' Something more energetic — like a dubstep feel where the beat hits the screen on every drop — would work better for a younger audience."

The tool made some changes, but the result still wasn't what I wanted. Interestingly, it also made a suggestion of its own: it offered to connect to a broader library and downloaded four music tracks, asking me to pick the one I liked best. The process was already becoming a back-and-forth, not just a one-way command.

Specific Feedback Made All the Difference

Once I moved from general feelings to concrete observations, things improved quickly.

"After the 19-second mark, the video becomes too static." → That section was restructured.
"Between seconds 5 and 12, we stay on the same shot for too long." → Scene durations were shortened.
"After the first five seconds, transitions should be more dynamic." → The opening pace picked up while the closing was preserved.

In each round I named exactly what I didn't like — with a timestamp or a clear description. After 5–6 rounds, the final version looked noticeably different and better than where we started.

The Real Takeaway

If you've spent weeks or months building a product, it's not realistic to expect a single prompt to produce a perfect promo video. You wouldn't tell a designer or photographer "make something great" in one sentence and walk away — you'd share references, give feedback, and go through revisions.

Working with AI on video is no different. The more specific your feedback, the more targeted the changes. "The music is bad" gives the tool almost nothing to work with. "The music loses energy after the 19-second mark" gives it a precise problem to solve. That difference in specificity is what separates a flat first draft from a version worth publishing.

What's Coming Next

I've noticed I'm spending a significant amount on WordPress plugins lately. I'm exploring how to replace them with AI-powered alternatives at no cost, and the next series of posts will be focused on exactly that.

If you've produced a promo video with AI and landed on something you're genuinely proud of, share it in the comments — seeing how others approach the iteration process is one of the best ways to get better at it.