
GPT Image 2.0
在 GPT Image 2.0 中比較兩個動漫角色參考表,包含角度、情緒和女學生細節
提示詞
目標:建立一張寬幅並排比較圖,呈現 夏目るり 作為學生會副會長版本的兩張動畫角色設定參考表;刻意以柔和的矩形模糊色塊遮住臉部。 畫布:超寬橫向畫布,尺寸約為 1200×429 px,白色背景,以一道細窄的垂直間隔分成兩個相等的面板。在左側面板上方中央加上一個小型綠色標籤,文字為 “GPT-image2”;在右側面板上方中央加上一個小型黃色標籤,文字為 “NanobananaPro”。使用細緻的藍綠色面板邊框。 左側面板配置:一張精簡的角色設計表,包含許多細緻的輔助線和小標籤。在左上方加入剛好 3 個大型角色轉面圖:正面、側面和背面。在右上方加入剛好 1 個標示為 “Eye” 的大型眼睛特寫,以及剛好 5 個標示為 “Mouth” 的小型嘴部範例。沿著底部以一列排列剛好 6 個半身表情/動作姿勢:正面中性姿勢、側面微笑姿勢、手靠近臉部的悲傷姿勢、講電話姿勢、單手吃三明治姿勢,以及頭髮放下的放鬆背面/側面姿勢。左側設定表稍微擁擠且不易閱讀,搭配較小的日式註解文字和細緻線條。 右側面板配置:更整潔、醒目的角色設定表。在左上角放置大型日文標題,文字必須完全如下:夏目るり 設定画。下方加上一行簡短的日文註記。在上方中央橫向呈現剛好 3 個大型頭部/肩部轉面圖,搭配水平輔助線和重複出現的 “1:1:1” 比例標記:正面、側面和背面。在右上方加入剛好 2 個局部特寫:一個帶有日文註解的綠色眼睛特寫,以及一個帶有日文註解的嘴部/嘴唇特寫。沿著右側面板底部加入剛好 6 個附有編號和日文說明的半身表情/動作姿勢:1 平靜的表情、2 嘴角微笑但眼神沒有笑意、3 苦惱的表情、4 神情銳利地查看手機、5 單手吃三明治,以及 6 放下頭髮。使用比左側更粗的輪廓線,並讓間距更清晰。 角色細節:動畫風格女學生,具有日本青少女氣質;柔和的棕色頭髮綁成高馬尾,並以藍色髮圈束起,略帶波浪的層次髮型,臉側有垂落的髮絲;白皙肌膚;在特寫中可見的大型淺綠色眼睛;灰色西裝外套制服、白襯衫、藍色蝴蝶結領帶、深色飾邊,展現整潔的學生會形象。表情姿勢應呈現沉著、嚴肅、略帶威嚴或禮貌開朗的感覺。在半身圖和轉面圖中,將所有主要臉部置於中央並以米色矩形模糊色塊遮住,同時保留頭髮、制服、雙手、三明治、手機和眼睛/嘴部特寫的清晰可見。 視覺風格:乾淨的數位動畫角色參考表、柔和的賽璐璐上色、低飽和色彩、細至中等粗細的黑色輪廓線、專業角色設定圖、日本設計表字體、許多小型註解線條、比較圖表式呈現。避免寫實風格,避免 3D 渲染,不要加入其他角色,除了上方兩個比較標籤外,不要加入標誌或浮水印。
Goal: Create a wide side-by-side comparison image of two anime character reference sheets for 夏目るり, a student council vice president version, with the faces deliberately obscured by soft rectangular blur blocks. Canvas: Extra-wide horizontal canvas, about 1200×429 px, white background, split into two equal panels by a thin vertical gap. Add a small green label centered above the left panel reading “GPT-image2” and a small yellow label centered above the right panel reading “NanobananaPro”. Use thin teal panel borders. Left panel layout: A compact character design sheet with many thin guide lines and small labels. Include exactly 3 large turnaround views across the upper left: front view, side view, and back view. To the upper right, include exactly 1 large eye close-up labeled “Eye” and exactly 5 small mouth samples labeled “Mouth”. Along the bottom, include exactly 6 bust expression/action poses in a row: neutral front pose, smiling side pose, sad pose with hand near face, talking on phone pose, eating a sandwich pose with one hand, and relaxed back/side pose with hair down. Keep the left sheet slightly dense and hard to read, with smaller Japanese-style annotation text and fine linework. Right panel layout: A cleaner, bolder character setting sheet. At the top left, place a large Japanese headline using this exact text: 夏目るり 設定画. Under it add a small note line in Japanese. Across the upper center, show exactly 3 large head/shoulder turnaround views with horizontal guide lines and repeated “1:1:1” proportion labels: front, side, and back. On the upper right, add exactly 2 feature close-ups: one green eye close-up with Japanese annotation and one mouth/lip close-up with Japanese annotation. Along the bottom of the right panel, include exactly 6 numbered bust expression/action poses with Japanese captions: 1 calm expression, 2 smiling but eyes not smiling, 3 troubled expression, 4 sharp expression checking a phone, 5 eating a sandwich with one hand, and 6 letting hair down. Use thicker outlines than the left panel and clearer spacing. Character details: Anime schoolgirl, teenage Japanese aesthetic, soft brown hair in a high ponytail tied with a blue scrunchie, slightly wavy layered hair with loose side strands; pale skin; large light green eyes where visible in close-ups; gray blazer uniform, white shirt, blue ribbon bow tie, dark trim, neat student council look. Expression poses should feel composed, serious, mildly stern, or politely cheerful. Obscure every main face in the bust and turnaround drawings with a centered beige blur rectangle, while leaving hair, uniform, hands, sandwiches, phones, and close-up eye/mouth studies visible. Visual style: Clean digital anime reference sheet, soft cel shading, muted colors, thin-to-medium black outlines, professional character setting art, Japanese design-sheet typography, many small annotation lines, comparison-chart presentation. Avoid photorealism, avoid 3D rendering, avoid extra characters, avoid logos or watermarks beyond the two top comparison labels.
- 提示詞為日文:直接貼上即可,模型能理解。
此提示詞的權利屬於作者 @Jp_Tnt。任何商業使用(推廣 AI 工具、聚合平台、Telegram 機器人)由你自行負責,mozgi.io 不承擔任何責任。 詳情 →











