Creating Video from Images: The Role of Generative AI and GPU

by Kael

Every time I add something new to my blog, it feels like I’m diving into uncharted territory. Unlike my husband, who keeps things steady, I use this space to experiment with things I’ve never tried before.

That’s probably why it feels all over the place—amateurish, even.

I have no formal training in music, I can’t draw particularly well, and I’ve neither worked in nor studied video production. Frankly, I know next to nothing about these fields, which means there’s a lot for me to learn.

The music I’ve shared is, embarrassingly, unmastered. Any professional would spot it immediately. That’s yet another skill I’ll have to pick up.

---

It took me just over a day to finish editing this video.

Tools I tried for the project include: Stable Video Diffusion, LivePortrait, Runway Gen-3, ComfyUI, Pinokio, CogVideoX, and a Python Conda environment. Of course, without my NVIDIA 3090 RTX, it would have been a much longer journey. Summed up, that’s pretty much the lineup.

I gave myself a hard time learning all of this, and even then, the quality isn’t quite where I’d hoped it would be. Still, I’ve decided to upload this as a record of my first draft, even if it’s a little embarrassing. Please don’t be too harsh… ha ha.

The images were created using MidJourney.

As I mentioned before, I wrote the lyrics myself. For the MidJourney characters, I specifically crafted prompts to create beautiful, non-human-like women. There are three main characters in total.

I aimed for a slight “noblewoman” vibe, but since this was my first time editing video, I borrowed Adobe Premiere Pro from Niel to piece everything together.