File size: 2,283 Bytes
db91c3a
f6825f0
c62597e
 
 
db91c3a
ad4c524
f7860cd
ad4c524
 
25f1683
 
ef23870
b6bb49a
621c060
42ef175
bf649a5
83851b8
 
 
 
98c3fe8
 
c3b274f
76259f3
3d81d83
98c3fe8
95e9765
5f9fd91
d321f5d
3b464f0
 
 
 
 
 
 
 
 
54be28f
34efe95
81a4486
73b980f
9b9fda0
c23ed38
e568cdb
8467018
2f01300
5deb74f
81a4486
37c4258
 
 
 
 
f43f506
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
---
license: cc0-1.0
tags:
- stable-diffusion
- text-to-image
---
# YuzuLemonTea Mix models ☕

List of my experimental merge models

- [Recommended Settings](#recommended-setteings)
- [YuzuLemonMilk](#yuzulemonmilk)
- [YuzuLemonChaiLatte](#yuzulemonchailatte)
- [YuzuGinger](#yuzuginger)

# important notice(Jan 15/23)

According to bbc-mc's note, there is a possibility of bug that some token(prompt) can be ignored, when merge with "add difference" option.
Milk and ChaiLatte models are now replaced with bug-fix ver.

https://note.com/bbcmc/n/n12c05bf109cc

# Recommended Setteings
VAE: "kl-f8-anime2" and "vae-ft-mse-840000-ema-pruned" are suitable

Steps: 20-30, Sampler: DPM++ SDE Karras or DPM++ 2M Karras, CFG scale: 8, Clip skip: 2, ENSD: 31377, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased),Denoising strength: 0.54~0.7

Negataive Prompt: (worst quality:2), (low quality:2),inaccurate limb,lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name
- (worst quality), (low quality) are adjustable between 1.4~2.0
- If you don't want 3DCG-ish paint, you can add (3d:0.8)~1.0 in Negative Prompt

# Sample prompt
4girls,(a 3d reader of:0.8) (teenage loli children:1.2), (wearing intricate casual camisole, cute hair ornament,crop jacket,hot pants, tighhigh:1.1),
shiny brown skin,
looking at viewer, (alluring smug:1.2),
dynamic angle,
(onomichi street:1.2),fisheye
<img src="https://i.imgur.com/2JiZwFU.jpg"  width="" height="1000">

# YuzuLemonMilk
Block merged model of Anything v3 and some real models.
Rather photo realistic.
Works fine with positive (realistic) and (photo realistic).
<img src="https://i.imgur.com/qYK8DKn.jpg"  width="" height="1000">


# YuzuLemonChaiLatte

Combination of a weight merge of ACertainModel and Anything-V3.0, and a block merge of several  realistic models.
Rather anime-ish style with realistic background.
<img src="https://i.imgur.com/Tim9BZK.jpg"  width="" height="1000">

# YuzuGinger

Add more anime models to YuzuLemonChaiLatte. Can be very anime looks.
<img src="https://i.imgur.com/4vc4HSL.jpg"  width="" height="1000">
<img src="https://i.imgur.com/96ka6hd.jpg"  width="" height="1000">