Ultrafast Super Generalized Tqing XL v1.5 TurboLcm

Be sure to read the introduction! Be sure to watch it~~!

My model can really produce anything. Accelerated models greatly improve drawing efficiency, and all styles You can control it with just a hint , real One model produces most styles

You can get whatever content you want by entering the corresponding style, eliminating the trouble of switching models, and not having to remember the characteristics of each model. The difficulty of use is greatly reduced, and the commercial value is extremely high

Attach some reminders for reference (Actually, there are more compatibility

, everyone is welcome to develop): High degree of realism Reminder: documentary photography

2.5D

style hint word:

pixar

2D style cues: manga, chibi, Makoto Shinkai, anime, sketching Negative Cues: (worst

quality, low quality, normal quality: 2),

can be used with

a sampler to recommend according to font size:

Euler a

DPM++ SDE Karras

DPM++ 2S a Karras DPM++ 2Sa

Can also be used, nothing else can use Iterative Steps (Steps

) At 4-8

Step by step has a very good performance

The Cue Guidance Factor (CFG Scale) is 1.5~2.5 Don't use 7 !!!

Nothing else to pay attention to, feel free to

fix a high score

, just don't exaggerate the image resolution (I have 1024*1536 and 2x 24G video memory will explode) The next step is model performance. I feel like I can really play whatever style you want. You can try out any style, but it's a mixture of quite a few

Let's take a look at the example image. Some of them have been returned to the image area, thank you very much

:

Pharmacist, Hui Ge, and Aunt Xue helped me test the

model and provide drawing ideas and design styles

Authentic style starting style:

Looking forward: looking at the camera, (pore: 1.2), Asia, ultimate details,

reverse：cartoon, painting, illustration, (worst quality, low quality, normal quality: 2),

Special note: Adding positive reminders to the AD plug-in can have unexpected effects