TL/DR: vanillasky.ai is like Lovable, but for short-form video. Prompt, get a short video, chat to refine it. I also created this demo to explain.
When I last posted, I was testing real estate Zillow-to-video. Before that, AI ad videos for small businesses. Before that, meme videos. I tried hard to make each one differentiated but the AI of the time wasn't far enough along. Whatever angle I picked, the big players like Canva were able to do it already or three other tools were doing roughly the same thing within weeks.
π οΈ What changed
A few months ago I started playing with the Remotion Claude skill, a public skill that lets you describe a video in Claude Code and Claude writes the React/Remotion code to produce it. It's been a hit; thousands of people have used it for explainer videos, code walkthroughs, kinetic typography demos, social clips. Watching a chat compose a video this way was the moment I realized the AI had caught up. What I thought would be impossible to build for years was suddenly possible.
But the more I used the Remotion skill, the clearer the gaps got:
That's when I knew what to build.
π What I built
VanillaSky is vibecoding for short-form video, the end-to-end product version of that loop:
Try it. No signup. Make a video about whatever you shipped this week, tell me what broke or felt off: vanillasky.ai
Thanks for sticking with me through the pivots π