Type | |
Stats | 977 |
Reviews | (136) |
Published | Jun 5, 2023 |
Base Model | |
Training | Steps: 3,200 Epochs: 20 |
Trigger Words | lissandra |
Hash | AutoV2 CCE1524251 |
LoRA trained on Lissandra from League of Legends
Coven Lissandra will be next
Initially, it was planned as a 2-in-1, but the helmet bleed into the other one even with Loha, so they will be separate
models used
anime: YabaL_Mix V3(image 7,12,13), MeinaAlter,
2,5D: DarkSun, HELLmix, GhostMix(image 2,3,8,9,10,11), MeinaUnreal - V3(image 1,5,6)
realistic: Paragon V1.0(image 4)
tags: lissandra, 1girl, blue skin
optional tags: helmet, long dress, pauldrons, braid, white hair, helmet over eyes
negative embeddings: embedding:bad-hands-5, embedding:ng_deepnegative_v1_75t, embedding:bad-artist, embedding:EasyNegative
Recommended settings:
Make sure you are aware on the usage instructions of LORA
VAE: vae-ft-mse-840000-ema-pruned (for realistic models & 2,5D), kl-f8-anime2, blessed2.vae (for anime models)
Sampler: DPM++ SDE Karras or DPM++ 2M Karras
Steps: 20 to 35
Clipskip: 2
Upscaler : img2img in comfyui using 4x-UltraSharp(for 2,5D & realistic), 4xAnimesharp(for anime)
CFG Scale : 5 to 9 (mostly 7)
LORA weight for txt2img: 0.5-1 (for the examples i mostly used 0.8)
Denoise strength for img2img: 0.4 to 0.7
for the pictures i used comfyui