
CFG: 3..6; Res: >> 768 (by shortest side). Steps:>40@800px and >24@1440x1080 w/Deep Shrink x2.
IMPORTANT: For better quality use higher resolution with Kohya's Deep Shrink (extension for A1111 WebUI or ComfyUI). Resolution of 768 shows medium quality with noticeable flaws.
This LoRa produces realistic image with Cavalier King Charles Spaniel dogs (Blenheim and Tricolor). If you want Tricolor CKSC variant, then add "black and white dog". If you want blenheim - add "brown and white dog".
This is the SWEETEST breed of dogs (by my preferences). I used the HiRes RAW photographs to train this LoRa. LoRa was trained in max resolution CivitAI allows to me.
I can train a Low Res version with the same Dataset if you request it in comments. For me the Hi Res version is more valuable.
PLEASE NOTE: If this LoRa at resolutions of 768 or lower and gives the low quality IT IS AN INTENDED behavior. Test it at higher resolutions with Deep Shrink before posting negative review.
PS: Deep Shrink can be easily installed to A1111 and ComfyUI
描述:
训练词语: NOBLEDOGS_CKCS
名称: NOBLE_DOGS_CKCS-000015.safetensors
大小 (KB): 74027
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success