剛好逛到一則 425.9K 觀看的提示詞, 整理成 WAI 工具可以直接使用的格式, 分享給大家,可以玩玩看: 這條提示詞很適合拆「動作場景連續性」的骨架:不是只寫有人打架,而是要把場景、角色數量、攻擊方向、鏡頭切換和收尾姿態全部鎖住。真正有價值的是「固定場景 + 主角身份 + 多個對手差異 + 每兩秒一個物理動作節點」這個結構。把酒館換成餐廳、倉庫、校園後巷、地鐵月台,或把武打換成產品追逐、遊戲預告、電影短片,都可以沿用同樣的節奏骨架。 WAI 工具試試: https://waigo.ai/share/DFwijpWQ ------ Segment 1: Tavern Fight Sequence GOAL: A black-clad female fighter defeats several assassins inside a crowded wooden tavern through fast, grounded martial-arts choreography with clear shot progression. CHARACTER main_fighter Female martial artist in fitted black travel clothes, dark leather belt, tied-back hair, agile grounded movement, realistic human speed, no supernatural effects, no identity drift across shots, no outfit color change. CHARACTER assassin_1 Masked tavern assassin wearing dark brown linen and leather bracers, lean build, short blade, different face and silhouette from all other assassins, no duplication. CHARACTER assassin_2 Masked tavern assassin wearing gray cloak and black gloves, heavier build, wooden staff, different face and silhouette from all other assassins, no duplication. CHARACTER assassin_3 Masked tavern assassin wearing faded red scarf and dark vest, medium build, dagger in reverse grip, different face and silhouette from all other assassins, no duplication. CHARACTER COUNT LOCK Exactly 4 combat characters total. No extra fighters, no duplicated bodies, no crowd members joining the fight. Strict count maintained entire scene duration. ENVIRONMENT Only one old wooden tavern interior for the entire segment: warm candlelight, rough tables, benches, hanging lanterns, wooden pillars, scattered cups, dust and smoke in the air, background patrons blurred and non-interactive. No scene change, no modern objects, no text signs, no additional weapons spawning. CAMERA Cinematic handheld close tracking, fast push-ins, whip pans between strikes, low-angle table-level tracking, brief impact emphasis only at contact moments. Keep bodies readable, grounded, and physically continuous.