【AI 繪圖】軟軟·阿比斯加德(Hololive EN)(SDXL-LoRA)

2024/11/19 38 0 繪圖作品 , VTuber , AI繪圖 ,

作品名稱:軟軟·阿比斯加德(Hololive EN)(SDXL-LoRA)
疊代步數:30 步
挑選張數:15 張
輸出尺寸:1024 x 768 px(4:3)
使用工具:Stable Diffusion web UI、Animagine XL V3.1+自行訓練3000步之LoRA模型
CFG Scale:7
提詞文字:((masterpiece), (amazing), (vivid), (elegant), (enchanting), (lustrous), (flawless), (best quality):1.2), 1girl, solo, (fuwawa abyssgard:1.2), __posture__, virtual youtuber, long hair, multicolored hair, streaked hair, blonde hair, blue hair, two side up, blue hairband, hairband, hair ornament, hairclip, bangs, hair between eyes, pink eyes, animal ears, animal ear fluff, jacket, collar, fox ears, long sleeves, dress, black jacket, white dress, blue nails, nail polish, frills, single thighhigh, fishnet thighhighs, open jacket, belt, short dress, cleavage cutout, clothing cutout, fox tail, tail, cleavage, large breasts, breasts, cowboy shot, __looking__, __expression__, __background__. <lora:fuwawa_abyssgard_XL:1>
__posture__:standing\nstanding\n(contrapposto:1.1)\n(contrapposto:1.2)\n(sitting:1.1)\n(sitting:1.2)
__looking__:looking at viewer\nlooking at viewer\nlooking at viewer\nlooking at viewer\nlooking at viewer\nlooking away, looking to the side\nlooking back
__expression__:smile/nsmile/nsmile/nsmile, open mouth/nsmile, closed eyes/nopen mouth/nshy/nlaughing/nangry/nevil/nlonely/ncrying
__background__:outdoors, starry sky, night sky, sunset\nnight, night sky, cloudy sky, meteor, planet, cloud, star (sky), (starry sky:1.2)\n(outdoors), day, blue sky, cloud, nature, tree\noutdoors, day, blue sky, cloud, water\noutdoors, day, grass\ncherry blossoms, tree, outdoors, shooting star, petals, cloud\noutdoors, city, cityscape, scenery, building

此次的目標是 VTuber「軟軟·阿比斯加德」(フワワ・アビスガード,Fuwawa Abyssgard),2023 年 7 月 26 日開始活動,隸屬於日本知名 VTuber 團體 hololive EN 中的「hololive English -降臨-」。擁有粉色瞳孔和米色長髮、挑染藍色,她以獨特的萌點著稱,如犬耳、犬尾、單腿襪、網襪、項圈等,展現出活潑與可愛並存的形象。身為魔界看門犬一員,她擔負著管理一切的角色,卻因一次離譜的惡作劇觸怒了眾神,最終被囚禁於大監獄「The Cell」。即便如此,她冷靜、可靠地照顧著她的雙胞胎妹妹 Mococo 和寵物 Pero,卻也常在外界眼中呈現出愛說話、愛玩、惹人喜愛的模樣。使用 SDXL 模型生成,並搭配自行訓練的 SDXL 專用 LoRA 模型,效果很不錯。總共產生 100 張,挑掉不合理跟瑕疵較多的有 15 張如下提供大家參考。















▼ 特效影片

贊助廣告 ‧ Sponsor advertisements

留言區 / Comments

萌芽論壇