Download
6 variants available
fp16 SafeTensor
CrossKemono-G2-RM-[FP16-NoEma]-VAEFIXED.safetensors
Half precision, best balance (pruned) • 2.3 GB
Verified: 3 years ago
SafeTensor
fp32
CrossKemono-G2-RM-[FP32-NoEma]-VAEFIXED.safetensors
Full precision, largest file (pruned)
Verified: 3 years ago
fp16
CrossKemono-G2-RM-[FP16+Ema]-VAEFIXED.safetensors
Half precision, best balance
Verified: 3 years ago
fp16
CrossKemono-G2-RM-[FP16-NoEma]-VAEFIXED.safetensors
Half precision, best balance (pruned)
Verified: 3 years ago
bf16
CrossKemono-G2-RM-[BF16-NoEma]-VAEFIXED.safetensors
BF16, good balance (pruned)
Verified: 3 years ago
1,0540 1 2 3 4 5 6 7 8 9,0 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
420 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
(95)
Oct 19, 2023

4050 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
3.6K0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
49.5K0 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K

NOTICE: I just realised the VAE was messed up and I can not fix it right now, so please use a VAE. Ps G2 will not have the problem
This is a recreation of CrossKemono but with a bit of my own personal flair.
The reason why I made this model was A, the original CrossKemono model was really old. B, I wanted to recreate a model which holds a special place in my heart while making it better, in my opinion.
Model Guide/stuff about the model
How this model was made:
This model was made through a pain staking process of merging over 60+ models then finding the right Lora weights while using Koyha_ss to extract the styles from the original CrossKemono model.
Model settings
Steps: 38, Sampler: DPM++ 2M SDE Heun Karras, CFG scale: 10,
Prompt: (best quality, highest quality, digital art, intricate, highres, 8k, anthropomorphic, furry, uploaded_on_e621:1.4), <lora:GoodHands-beta2:1>,
Negative prompt: (worst quality, low quality, normal quality:1.6), watermark, artist logo, signature, patreon, text, simple background, sepia, simple eyes, empty eyes, no pupils, multi limb, multi arm, multi leg, transformation, blurred background, blurred foreground, sketch, (bad quality, deformed, bad eyes, lowres), boring_e621_fluffyrock_v4:1.5, boring_e621_v4:1.5, FastNegativeV2:1.5, NGH:1.5, <lora:neg4all_bdsqlsz_V3.5:1>,Prompting
Prompting for this model is nearly Identical to CrossKemono But It has a Lot of concept recognition becuase of it being merged with Fluffyrock, E6lion and P.A.W.F.E.C.T-Alpha.
