I haven’t updated the blog in a while but I’ve been working on some projects. None of them are finished though but the one I’m describing here is “live” in the sense that people can interact with it.

I am intrigued by the possibility of creating “viral” videos using AI. There’s so much interesting knowledge in these LLMs like o3. They have such fascinating stories to tell if you let them, by prompting them in the right way.

So I decided to create videos using the o3 model. I ask it to generate a series of “scenes” by describing the initial frame and the motion. I then use a video generator to extend those frames into 5-10 second clips. I stitch them together, add narration (also by o3), I use Elevenlabs to generate the audio and I overlay subtitles.

That’s basically it. The results is quite cool. I just type what I want a video about and 1 minute later it just spits out the video.

Why I think it might work: I feel like there is a gap on TikTok, Instagram and Youtube Shorts for educational videos that are put in a “viral” jacket. So it looks like the typical brainless slop, and it’s entertainning to watch, but it’s actually teaching you something.

I mean, that would be quite cool if it works right?

Why I think it might not work: the videos might still be too generic and too “AI”-like. I don’t know if the tech is already quite there. + People get used to AI stuff so incredibly fast that they might just think it’s super boring once they see a few of them.

So yeah let’s see whether this pans out or not. I do like experimenting with it but the tech might not be there yet.

Here’s an example of one of my AI educational shorts:

And here’s another one: