Type | |
Stats | 63 217 |
Reviews | (8) |
Published | Mar 12, 2024 |
Base Model | |
Training | Steps: 3,600 Epochs: 10 |
Usage Tips | Strength: 1 |
Trigger Words | kyle_hyde |
Hash | AutoV2 3912410891 |
Character model of Kyle Hyde from Hotel Dusk and Last Window.
Trigger word: kyle_hyde
There isn't much art of him so I'm not sure how good this LoRa ended up. Not to mention images have a pencil drawing style to them. Will upload previous Epochs if necessary.
For version E16 (SD 1.5): I used 16 images X 10 repeats X 16 epochs / 2 batch sizes = 1280 steps.
The dataset used for training this model included some sketches / pencil drawings and a few images of lower resolution.
For version E20 (SD 1.5): I used 10 images X 10 repeats X 20 epochs / 2 batch sizes = 1000 steps.
In this version I used less images of a better quality, and less black-and-white drawings. Maybe this will improve results.
After some testing, both SD 1.5. versions seem to be working fairly well with the help of some popular checkpoints and more detailed prompting. Mature Male Mix gives some good results from what you can see in the showcase images, but other popular checkpoints also work well.
SDXL version: trained directly on civitai, with a dataset containing 20 images and updated tags: 20 images X 10 repeats X 10 epochs / 2 batch sizes = 1000 steps. I need to do more testing but maybe because I included more images (including some of lower resolution) or because I modified the tags, the model seems less reliable in producing high quality results. Plus instead of realistic images it still keeps the pencil drawing style. Let me know if you can manage to make it work well.
SDXL V2 version: trained with Tensor Art's online training, with Animagine XL. Updated tags, used only 18 upscaled images of higher resolution: 18 images X 20 repeats, X 10th epoch = 3.600 steps.
This version might be a little under-baked as the character doesn't really look that much as Kyle Hyde sometimes, but it might be better for realistic images. From initial tests, they don't seem to create good results for anime images.
General tag structure for this version: kyle_hyde, <<type of image: pencil drawing / pencil sketch>>, <<optional: drawing in thick lines, monochrome / grayscale>>, 1man, <<face portrait / upper body / full body>>, <<perspective (optional): side view>>, <<actions: looking at viewer / smiling / thinking / holding phone / closed eyes / holding gun / sitting on couch / holding and reading book>>, <<clothes: formal black suit / blue jacket, white collared shirt, black necktie>>, <<expression: frown / serious face / surprised expression>>, hands in pockets, <<background: white background / dark city street / indoor background / blurry background of desert>>