Hello,
Everyone who follows my profile knows that I like to make quite a few weird Lora models. They're not always entirely predictable, but in return, they often offer a unique atmosphere and an interesting artistic expression.
This time is no different; the model was supposed to create frames with a repeated image inside them. It's quite unpredictable (often creating illustrations of men or women holding the frame, sometimes completely ignoring the prompt). So, the flexibility of the model is quite specific; sometimes it does something completely different than what we wanted. We have some influence on this by changing its weight (I tested from 0.6 to 1.4) and selecting different base models.
Feel free to download and have successful experiments! 😊