Type | |
Stats | 47 |
Reviews | (4) |
Published | Sep 9, 2024 |
Base Model | |
Training | Steps: 1,200 |
Usage Tips | Clip Skip: 1 |
Trigger Words | orban |
Hash | AutoV2 1CF61E7D23 |
This LoRA is designed to recreate a likeness of a 61-year-old slightly overweight Viktor Orbán. It works optimally with the "orban" trigger word, and the phrase "61-year-old slightly overweight Viktor Orbán" best reflects his body type and facial features.
In my setup, the following parameters consistently produced the best results:
Scheduler: Normal
Steps: 25
Denoise: 1.00
Guidance: 3.5 (via FluxGuidance)
Sampler: Euler
Dual CLIP Loader:
clip_name1: clip_l.safetensors
clip_name2: t5-v1_1-xxl-encoder-Q8_0.gguf
I used the flux1-dev-Q8_0.gguf model with a LoRA strength setting of 1.00 to 1.10. The image dimensions were set at 768x1024 pixels to achieve an optimal balance between output quality and processing speed.
Please note that the dataset was cherry-picked for the best results, so you may encounter some goofy or unexpected outputs occasionally. However, in most cases, the results are great and reliably capture Orbán's likeness.
This setup has proven effective for high-quality image generation, and adjustments can be made to suit different needs or scenarios.