
Create a Girl in Stability Matrix with SDXL
Good morning everyone, I hope you can help me with some suggestions. Just as a premise, I’ve only recently started getting into this world, and I had the idea of creating an AI model/influencer just for fun and personal enjoyment, then opening a social media page for her.
Over the past few weeks I’ve been getting help from Claude and Gemini, which pointed me toward creating a local “lab” using Stability Matrix and running models through Stable Diffusion WebUI Forge. The models I’m using are Juggernaut XL Ragnarok and RealVisXL V5 Lightning (BakedVAE).
After setting everything up, I tried creating a coherent image dataset so I could later train a LoRA with Kohya_SS.
So what’s the problem? I just can’t understand why I’m completely unable to create a consistent dataset.
My plan was to make:
- 10 close-up portraits
- 10 half-body shots
- 10 full-body shots
- plus various images from different angles
I managed to get a close-up image of the following girl [photo], and I saved the seed. But even though I keep the same characteristics from the original prompt — basically trying to keep the same girl but from different angles — the girl is NEVER the same.
I’m wondering what I’m doing wrong.
had actually managed to get a first set of a similar-looking girl using Google Banana, and paradoxically I don’t even know why, but the consistency between one image and another was almost perfect without having to go crazy trying to achieve it.
However, I wanted to create one inside Stability Matrix because, understandably, I’m not sure whether Google’s policies allow the use of images created with Nano Banana.
Can you help me with suggestions?