Type | |
Stats | 73 |
Reviews | (12) |
Published | Jan 14, 2024 |
Base Model | |
Training | Steps: 1,400 Epochs: 1,400 |
Usage Tips | Strength: 0.5 |
Trigger Words | koitsu (artstyle) |
Hash | AutoV2 75839E4336 |
This model is intended as a shader not a character specific, or standalone character generator model. The fle is a SDXL Lora but was trained using SDv1.5 using a google colab project. So random tip.
1: Use furryrockv70 as the checkpoint model. 2nd, set both my lora and a character lora to be on equal values. My lora file might be 0.7 while a character specific lora is 1.0. Next set the steps to 48, use either Euler or DPM ++ 2M Karras for best results. Then render your image in bulk. I recommend 4-8 images. my Koitsu models is intended for shading and video game texture packs for local hardware users.
Changelog:
1.0 uses 1 image reffernce.
1.1 uses 20 image refferces, The wolf link render uses both my koitsu (artstyle), and link (wolf form) v 1.3 model Anthro Wolf Link - TOTK Wolf Link beta | Stable Diffusion LoRA | Civitai
I used furryrockv70 as the chkpt file. Otherwise, I used nmkd stable diffussion to render the art after 15 hours T.T I swear I need a new good computer lol
version 1.2 was a style changed.
Version 1.3 was an attempt to make a better model based on one image similar to v 1.1 but, to correct eye issues.