General mix's successor animation model!
Feel free to propose any suggestions or opinions in 820725763 (QQ group) during use!
It integrates multiple models of both realist and animated styles, adopts MBW hierarchical merging, and incorporates the self-trained Earth OL style, capable of producing three schools of D3/2.5D/2D.
Unlike the Everything Furnace, the images it produces are more stable and support extreme resolution magnification.
Guaranteed that 3840×3840's i2i and 2560×2560's t2i won't have a lot of crashes when opening Hires. fix.
Recommended parameters are similar to Animated models of General mix's:
DPM++ 2M SDE Karras Sampling method,
Don't exceed 768px for direct image resolution,
It is recommended to add Bad hands-5 to reduce the probability of bad hands,
Using R-ESRGAN 4x+ Anime6B, i2i can give this model an effect superior to Hires. fix,
For magnification, tile VAE is recommended instead of direct magnification, can reach up to 5120×5120 with CNT without color stratification,
Image aspect ratios shouldn't exceed 16:9 or 9:16, else there will be severe crashes.
Improvements:
Fixed ugly mouth appearance, though it may still look bad in some cases,
The model is more likely to hide hands,
Changed the trigger word, now to make realistic images only need Photos, real ones two trigger words, to do animation, you need to use Anime, Illustration for reinforcement,
Improved NSFW character representation, plus realist content,
More sensitive to scene tags, significantly improved details, clearer distant content.
Negative Optimization:
The ability of animated style sfw representation is worse, and the scene doesn't have the gloss of Animated models of General Mix,
Always wants to draw people on landscape pictures,
Certain overfitting is present in the scenes.
Reduced ability to handle scenes with people.
All info on non-original combined materials are as follows:
1.https://civarchive.com/api/download/models/109115
ICBINP - "I Can't Believe It's Not Photography"
by：residentchiefnzhttps://civarchive.com/user/residentchiefnz/models
2.https://civarchive.com/api/download/models/111969
ouka_star
By：ouka_studiohttps://civarchive.com/user/ouka_studio/models
3.https://civarchive.com/api/download/models/67206
ChickMixFlat
by：xxxholichttps://civarchive.com/user/xxxholic/models
4.https://civarchive.com/api/download/models/100675
MeinaHentai
by：Meinahttps://civarchive.com/user/Meina/models
5.https://civarchive.com/api/download/models/108772
SakushiMix-Hentai
by：Novamindhttps://civarchive.com/user/Novamind/models
6.https://civarchive.com/api/download/models/119667
Veela
by：CyberMhttps://civarchive.com/user/CyberM/models
7.https://civarchive.com/api/download/models/96764
Stoked Illustration
Description
This will be my last SD1.5 model, leading the world in all things is not for nothing. It can perfectly allow you to reach a degree of "one model in hand, the world is mine". Please read the document below carefully to better control it.
Recommend Negatively
(Pablo Picasso:1.2),(Salvador Dali:1.2),(low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2),lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, (bad-hands-5),(bad-hands-3)
Recommended positive style tags
Add ((Anime)),((Game CG)),((Illustration)) at the start of the tag
Add ((Paintings)),((imagery)) at the end of the tag
This will result in less 'AI flavored' animated pictures. When the tag is too long, it may not work properly.
Add ((Real World)),((Photographs)),((Live Action)) at the start of the tag
Add ((3D modeling)),((3D style)),((three-dimensionality)) at the end of the tag
This will result in more realistic background images, the effect on characters is not very obvious.
Recommended output parameters
Sampling
DPM++ 2M SDE Karras
50-70 steps
CFG7-12
Recommended image ratio
Horizontal 16:9, 9:16
Vertical 8:5, 5:8
Or square 1:1
Do not use 4:3 or 3:4, otherwise, the face is very likely to mess up.
Resolution within 768
60+ steps can automatically fix the face without additional facial repair enabled.
HD restoration suggestion
Magnification 1.5-3 times
Restoration method 4x-AnimeSharp
40-60 steps
Redraw amplitude 0.45-0.6
Regular drafts can usually achieve effects better than HD restoration.
Magnification 2-4 times
Restoration method 4x-AnimeSharp
50-70 steps
Redraw amplitude 0.3-0.45
Other specifications
This model doesn't need to add a long section of quality TAGs, if you need style TAGs, use the ones I recommended.
This model doesn't need AVE, it will lead to over-color fitting.
Red Moon wrote in July 28, 2023
Used for merging materials:
Animated models of General mix
https://civitai.com/api/download/models/107617
I Can't Believe It's Not Photography
https://civitai.com/api/download/models/121557
ouka star
https://civitai.com/api/download/models/111969
ChickMixFlat
https://civitai.com/api/download/models/67206
MeinaHentai
https://civitai.com/api/download/models/120702
SakushiMix-Hentai
https://civitai.com/api/download/models/127453
Veela
https://civitai.com/api/download/models/119667
Stoked Illustration
https://civitai.com/api/download/models/96764
Clamp xxxholic Style
https://civitai.com/api/download/models/63234
Special thanks：
墨幽
吃肥皂泡泡
Vankary
追光者
无味先生
哈尔法狗
ℳ๓₯㎕
南颜
栀子花 白花瓣
黄金晨曦
奥斯瓦尔多
丢了猫のBOY
微信实名上网
reluxion
FAQ
Comments (1)
加油，今日返图