Okay, so today I wanted to mess around with making a little something about Linda Valenzuela. I’d seen some cool AI-generated images of her floating around, and I figured, why not try to make my own?
First, I needed to figure out what tools to use. I’m no expert in this stuff, so I just went with what I’ve seen other people use before. After I installed and tried Stable Diffusion for image generation, along with some character LoRA , this is my first time to use them.
Getting Started with Image Generation
I fired up Stable Diffusion. Honestly, the interface was a bit overwhelming at first. So many buttons and sliders! I basically ignored most of them and focused on the text prompt area.
My initial prompts were super basic. Stuff like, “Linda Valenzuela, portrait,” just to see what would happen. The results? Uh… let’s just say they were interesting. Some of them looked nothing like her, and others were just plain weird. It was clear I needed to be more specific.
Refining the Prompts
I started adding more details to my prompts. I tried things like:
- “Linda Valenzuela, 80’s style, photograph”
- “Linda Valenzuela, low-angle shot, dramatic lighting”
- “Linda Valenzuela, smiling, casual pose”
I also started playing around with some of the settings, like the “sampling steps” and “CFG scale”. I don’t really understand what they all do, but I just tweaked them until I started getting images that looked a little better.
I also tried to put some Negative prompt to avoid some bad result. Sometime the result turned into wired, with extra fingers, bad face looking, or some random objects.
Experimenting with Character LoRAs
I used the models to see my prompt working or not, because these models are trained by others, and looks more familar to me.
Iterating and Tweaking
This whole process was a lot of trial and error. I probably generated hundreds of images, most of which were pretty terrible. But, slowly but surely, I started getting some that I actually liked. I’d save the ones that were close, then tweak the prompt or settings a bit more and try again.
It was kind of addictive, actually. Like a slot machine, but instead of money, you’re gambling with your time and hoping for a cool picture.
The Final Results (Well, Some of Them)
I’m still no expert, but I managed to create a few images that I think are pretty decent. They’re not perfect, but they definitely capture some of the essence of Linda Valenzuela, I tried to get all of them close to 80’s Style.
It was a fun little experiment, and I definitely learned a lot. Maybe I’ll try this again with a different subject sometime. I also want to try video generation.
