Experimenting With AI Video

Sorry I’ve been so quiet this week. Feeling a profound need to pull away from social media and lean hard into creative work. Hence the fiction last Friday, and now this week’s experiments with AI video.

Well, whatever AI video I can do for free. Decided to make an animated version of one of my favorite poems...Shelly's "Ozymandias."

This has taken two days of work to get done. Lots of that was learning, generating, re-generating, waiting.

We're getting to a place where this is more usable. But the workflow and quality still leaves a lot to be desired.

Everything in the video is AI generated EXCEPT the last scene of the desert. I was out of credits. I just downloaded some stock footage.

The voiceover is just me.

The amount of stuff I had to render into images to later turn into video was ridiculous. So many iterations to get the right look and field. The compilation image below is just my Midjourney renders for the statue's head. And those just from today. I honestly have no idea how many hours it took to render everything.

Glad to have learned, but this needs to be streamlined and expedited. A lot.

Tools used:

• Midjourney

• ChatGPT

• Hedra

• Sora

• Runway

• Artlist

• Adobe Premiere

• Adobe Audition

Still, I do think things are getting much better for individual creators. Here’s something I saw online today that I was pretty blown away by. A trailer for a non-existent film made by a single creator. Extremely well done. Way beyond what I can do for now (you’ve got to pay for access to these platforms, and they are not cheap):