这是v2.0的

This is v2.0

大多数1.5大模型应该都可以用于生图。

Most 1.5 large models should be able to be used for graph generation.

欢迎评论反馈，渴求返图和使用。

Welcome to comment and feedback, eager to return to the image and use.

lora权重：0.9-1.2，用1即可。

LoRa weight: 0.9-1.2, use 1.

特别的是clip，二次元用2，现实风用非2，如1。

In particular, the anime clip uses 2, while the real wind uses non 2, such as 1.

• 触发词：logo，icon，

• Trigger words: logo, icon,

可以单用，也可以双用。我建议用其中一个就好。

一般仅使用icon,(circle frame),即可

It can be used alone or in combination. I suggest using one of them.

Usually, only used: icon, (circle frame), which is sufficient.

• 训练用词：circle frame，

• Training words: circle frame,

• 仅需一个词即可召唤，logo可以单用，不需要加（circle frame），。

• Only one word is needed to summon, and the logo can be used alone without the need for a (circle frame),.

• icon生成人物最好加上（circle frame），这会更加可控。

• It is best to add a circle frame to generate characters for icons, which will be more controllable.

• 当描述一个东西，提示词太多的时候，过于详尽的概念可能会偏移。

• When describing something with too many prompting words, overly detailed concepts may shift.

可以加强prompt权重如：（icon），（circle frame：1.3）

The prompt weight can be strengthened, such as: (icon), (circle frame: 1.3)

• logo主要用于生成标识标志，带有English text，可生成动物和无人图标。

• The logo is mainly used to generate identification marks, with English text, and can generate animal and unmanned icons.

不要English text可在负面提示词加：（English text），（artist name），

Do not use English text. You can add: (English text), (artist name) to negative prompts,

• icon最好加上（circle frame），无English text，可生成人物，动物，其它等图标。

• It is best to add (circle frame) to the icon, without English text, which can generate icons such as characters, animals, and others.

icon更多使用大模型的概念生图。

Icons use more conceptual diagrams with large models.

• 与其他lora结合，该lora权重1，另一个lora权重在0.7表现还不错。

• Combined with other Loras, this Lora has a weight of 1, while the other Lora performs well at a weight of 0.7.

• 可能有用的词：square frame,

• Possible useful words: square frame,

• 当你只想要圆圈，不想要框：(circle frame)—>(circle)

• When you only want circle and don't want frame: (circle frame) ->(circle)

是的，现实风不容易生成人物图标，当不同提示词详尽地描述时，很容易概念偏移，人的概念会更强，而不是图标的概念，logo与icon是不同概念，所以我也做出来区分。

与原来相似，我加入了frame进行训练，当生成人物可以加上（circle frame），在某种程度达成了生成人物图标的目的。

Yes, realistic style is not easy to generate character icons. When different prompt words are described in detail, it is easy for concepts to shift, and human concepts will be stronger than icon concepts. Logo and icon are different concepts, so I have also made a distinction.

Similar to before, I added a frame for training, and when generating characters, (circle frame) can be added, to some extent achieving the goal of generating character icons.