一、先寫角色錨句，再寫每一鏡
AI 影片生成工具每一次生成都是獨立的，它不記得上一鏡的人長什麼樣子。所以要先把人物長相、年紀、髮型、衣著、場景、光線、畫面比例寫成一段固定文字，每一鏡的提示詞都把它放在最前面。少了這一步，15 個鏡頭會長出 15 個不同的人。
提示詞庫 Prompt library
15 條提示詞，全部公開
這裡沒有付費牆，沒有留 email 才給。每一條提示詞都對照得到成品的哪一鏡，複製鈕會自動幫你把角色錨句接在前面。
01
怎麼寫這種分鏡提示詞
二、一鏡一條，不要把整段劇情塞進一條
一條提示詞只描述一個畫面。把「他被砸了筆、蹲下撿、然後決定畫畫」寫成一條，生成出來會是三個動作混在一起的糊畫面。拆成三鏡，每鏡各一條。
三、寫看得到的東西，不要寫看不到的
「他心裡很不甘」生成不出來，「特寫他的眼睛：沒有情緒」可以。把情緒翻譯成鏡頭語言與可見的動作，這是分鏡提示詞跟一般文字提示詞最大的差別。
四、不要寫秒數
寫秒數會讓工具把時間也當成畫面內容去理解，構圖反而失焦。節奏交給剪輯，提示詞只負責描述這一鏡要看到什麼。
五、直式 9:16 要寫進錨句
短劇是直式的，但多數工具預設橫式。畫面比例要寫在錨句裡，每一鏡才會一致，不然會拿到一半橫一半直的素材。
02
全部提示詞
All prompts鏡頭尺寸是直接讀提示詞裡實際出現的英文詞判讀出來的（例如 extreme close-up、camera pulls back），不是人工分類，也不是猜的。判讀不出來的歸在未標註。
第 1 集：全工坊都笑我只配洗筆
Renaissance painter's workshop in Florence, 1470s: a young apprentice Leonardo da Vinci (about 20, shoulder-length curly hair, keen gentle eyes, plain linen tunic) among easels, pigment jars and brushes; warm candlelight or morning light, painterly chiaroscuro, vertical 9:16.
A bundle of dirty paintbrushes hurled at the young apprentice's head, brushes scattering across the floor; a sneering senior apprentice looms, others laugh.
The youth kneels and gathers the brushes one by one; extreme close-up of his calm, emotionless eyes.
Night, empty workshop, a single candle moving through darkness; the youth stops before the master's large unfinished painting; a blank corner of canvas.
Extreme close-up: the brush tip lands, fluid confident strokes, a small angel taking shape on the canvas.
Final stroke; macro close-up: the painted angel's eyelashes flutter once; the youth leans in and raises a finger to his lips: shh.
Morning uproar: apprentices crowd the big painting; the senior grabs the youth by the collar, shouting in his face.
The old master enters, the room falls silent; he walks to the painting and stares at the angel; slow push-in on his weathered face.
He raises his hand, apprentices flinch, the youth shuts his eyes; close-up: he snaps his own brush in half with both hands.
A dignified steward in Medici livery steps through the doorway, backlit; everyone freezes mid-motion.
The senior apprentice steps forward eagerly, hand pressed to his chest, chin high.
The steward smiles and places a brush into the senior's hand, gesturing toward a blank wall.
The senior's face collapses in stages, sweat beading; the brush clatters to the floor; the youth bends down and picks it up.
Close-up on the wall: three swift strokes reveal an angel's face; camera pulls back to a stunned, dead-silent crowd.
The steward bows deeply to the youth; walking out with his satchel, the youth presses the old washing brush into the senior's hands with a faint smile.
Carriage interior: an open notebook of mirror-written script puzzles the steward; a wax-sealed parchment is handed over; the youth breaks the seal, his eyes widen in shock; hard cut to black.
這個鏡頭尺寸目前沒有提示詞。
03
常見問題
FAQ什麼是角色錨句？為什麼每一鏡都要加？
角色錨句是一段固定描述人物長相、場景與畫面風格的文字，放在每一鏡提示詞的最前面。AI 影片生成工具每次生成都是獨立的，不加錨句的話，同一個角色在 15 個鏡頭裡會長成 15 個人。錨句是接龍能接得起來的技術前提。
提示詞為什麼用英文寫？
目前主流的 AI 影片生成工具對英文提示詞的理解仍然比較穩定，而且英文提示詞在不同工具之間的可移植性比較好。劇本本身是中文，只有提示詞用英文。
為什麼提示詞裡不寫秒數？
寫秒數會讓生成工具把時間也當成畫面內容去理解，反而讓構圖失焦。節奏交給剪輯處理，提示詞只負責描述這一鏡要看到什麼。
這些提示詞可以直接商用嗎？
可以。劇本與提示詞採 CC BY 4.0，署名即可自由使用、改作、商用。你拿去開課、做商案、改寫成別的故事都可以，只要標明出處。影片本身的著作權屬各集創作者本人。