V1.5Good at 0.8trigger word : lennigan outfitcomponent trigger word : red ribbon, white pantyhose, shoulder armor, elbow guards, red bow,maid headdressComplete outfit trigger words (single-character scenes) : lennigan outfit, black dress, armored dress, red ribbon, white pantyhose, shoulder armor, elbow guards, long sleeves, red bow, white frills, maid headdress,这个版本我使用了新的打标策略,并且进行了40轮的训练,它现在需要''总触发词和部件触发词''的形式生成服装,相对于V1.0,除了服装细节更还原,生成多人场景不那么容易把服装特征串到异性上了This version uses a new captioning strategy and was trained for 40 epochs. It now requires the format of “main trigger word + component trigger words” to generate the armor.Compared to V1.0, not only are the armor details reproduced more accurately, but in multi-character scenes, it is also less likely to transfer the armor features onto characters of the wrong gender.V1抱歉,我犯了个低级错误,我把触发词设置成了Lennigan_Outfot而不是Lennigan_Outfit,不过从结果上看这并不影响生成.Sorry, I made a basic mistake. I set the trigger word to <code>Lennigan_Outfot</code> instead of <code>Lennigan_Outfit</code>. However, based on the results, this does not seem to affect the generation.Trigger Word: Lennigan_OutfotWithout Tutbo Lora : Good at 0.7-1.0With Turbo Lora :Good at 0.6

Monster Hunter : Lennigan Outfit

Good at 0.8trigger word : lennigan outfitcomponent trigger word : red ribbon, white pantyhose, shoulder armor, elbow guards, red bow,maid headdressComplete outfit trigger words (single-character scenes) : lennigan outfit, black dress, armored dress, red ribbon, white pantyhose, shoulder armor, elbow guards, long sleeves, red bow, white frills, maid headdress,这个版本我使用了新的打标策略,并且进行了40轮的训练,它现在需要''总触发词和部件触发词''的形式生成服装,相对于V1.0,除了服装细节更还原,生成多人场景不那么容易把服装特征串到异性上了This version uses a new captioning strategy and was trained for 40 epochs. It now requires the format of “main trigger word + component trigger words” to generate the armor.Compared to V1.0, not only are the armor details reproduced more accurately, but in multi-character scenes, it is also less likely to transfer the armor features onto characters of the wrong gender.

Files with SHA256: 281a02e279b54ab604ab464924135d8c91cf0cb1b600a0f14c587fe3c98e7f73