SDXL游戏风_V1丨漠北此模型是针对游戏海报的训练，旨在为了游戏海报工作者提供创意参考。欢迎大家多多使用返图，多多提供意见反馈，您的意见是我更新下去的动力(●°u°●)推荐参数：采样方法：Euler aCFG:4~6   5效果最好Steps:20~30触发词：Game style,WANGZHE,（用其中1个也可以）建议开启高清修复建议开启AD局限性训练集尺寸较为单一,建议尺寸1334X704px，全部为横版，过高或者过低可能导致效果变差训练集只有600张，在泛化性上有较大局限性，建议使用强化训练的TAG，模型只在画风上进行了加强，脸部和手部及动作任然有较大概率崩坏提示词强化训练：weapon（武器），gloves（手套），holding weapon（持有武器），white hair（白色的头发），chinese clothes（中国衣服），animal ears（动物耳朵），looking at viewer（看着观众），glowing（发光的），sword（剑），hair ornament（发饰），ponytail（马尾辫），dress（裙子），或者大家可以使用游戏海报进行反推获得提示词，基础模型：Animagine XL v3.0-base

SDXL游戏风_V1丨漠北

sdxlV1_v10.safetensors

v1.0

SDXL Game Style_V1丨Mo bei

SDXL Game Style_V1|MoBei
This model is designed for training on game posters, aiming to provide creative references for game poster creators.
Feel free to use the generated images and provide feedback. Your feedback is the driving force for my future updates (●°u°●)
 
Recommended Parameters:
Sampling method: Euler a
CFG: 4~6 &nbsp; 5 works best
Steps: 20~30
Trigger words: Game style, WANGZHE, (using either one is fine)
High-definition restoration recommended
AD recommended to be turned on
 
Limitations:
The training set size is relatively uniform, recommended size is 1334X704px, all in landscape format. Deviating too much from these dimensions may result in decreased effectiveness.
With only 600 images in the training set, there are significant limitations in generalization. It is recommended to use reinforcement training tags.
The model has been enhanced only in terms of art style; there is still a significant probability of distortion in facial features, hands, and movements.
 
 
Enhanced Training Cue Words:
weapon, gloves, holding weapon, white hair, Chinese clothes, animal ears, looking at viewer, glowing, sword, hair ornament, ponytail, dress,
 
Alternatively, you can reverse deduce cue words from game posters.
 
Base model: Animagine XL v3.0-base

Files with SHA256: 3d1b6c2157772caf07d98a51e6ad0882d466a5b34ad1f8a0db59f61c17ad76fd