少帅｜汉卿啊！你的感情在哪啊
制作资料包含 16 张去重后的分镜原图，以及 36 条原始提示词。
生图提示词 17 条（含 1 条男主首版历史记录）；视频提示词 19 条（含 3 条历史对比记录）。
分镜按素材内容分为感情问答、全国覆盖、两省一市、女主回怼，组内按原文件编号排列；不代表成片剪辑时间线。
视频提示词与图片的关系来自原始输入参考和图片 SHA-256 核验，不代表每张图都有独立视频提示词，也不代表这些版本全部用于最终成片。
historical=true 为早期或历史对比版本；其余版本沿用原资料名称，不作最终成片版本认定。
prompts 文件夹中的 TXT 保存完整提示词原文；manifest.json 提供类型、版本、关联分镜和 SHA-256 校验值。
本包不包含成片视频；成片原文件在页面单独下载。分镜原图与成片原文件均未重新压缩。


================================================================
prompt-001｜女主低头欲言
类型：生图　版本：原始四张分镜　历史记录：否
来源文件：提示词与参考记录.json
关联分镜：storyboard-01

生成一张16:9横向独立电影分镜静帧，photorealistic 真人电影质感，无文字、无边框、不是拼图。镜头01「低头欲言」，中近景裁切至胸口下方，画面只出现一位女子。参考输入依次为：第1张是唯一的女主身份、低发髻和服装参考，必须保持她的脸型、眼形、鼻唇、黑发低髻、素灰簪和银灰墨染交领衣；第2张仅负责镜头构图与低头表演，不继承其中演员、民国红衣、耳饰与彩色室内；第3张只负责白石矿脉、黑木和墨染织物的场景材质，不继承其中任何人物的身份，不加入第二个人。
保持参考视频女主正反打结构：女主坐在黑木低桌一侧，胸肩位于画面左中部，身体向画面右方，脸呈右向三分之二侧面，视线落在右下方，头轻垂，唇略分开但未展开说话，内敛犹疑，像将情感问题轻声说出前的一瞬。不是对镜头摆拍，没有整理袖子或抬手动作，双手在画面下缘外。面部占画高约四成，头顶保留余量，右方留出视线空间，85mm级中长焦自然透视。
女主面容以第1张为准：清秀自然椭圆轮廓、略长杏眼、细直鼻梁、自然淡玫瑰色唇，肌肤保留微小真实肤理、极淡暖色与连续的颧颊鼻侧体积。黑发在脑后低髻，灰簪、两缕鬓发位置延续。银灰层叠交领、白内领和可见肩袖上的大片浓淡墨染，染纹贴附织物顺褶垂坠。
背景是同一古典庭院廊下：明亮洁净白石墙上的连续宽灰矿脉，有尺度与方向变化；画面左边缘一根少量黑木柱，近景不加入池、伞、松或额外茶具。庭院阴天漫射光从右前方柔和塑脸，眼窝有浅填充，后方白石保留层次和纹理；白灰黑来自材料本色，皮肤只有极淡暖色。背景稍虚而白石宽脉依然可辨。不要换成原视频女演员，不要高髻或长垂耳饰，不要普通灰水泥、纸纹水墨画、粗粒滤镜、塑料磨皮。


================================================================
prompt-002｜男主沉默凝视（首版）
类型：生图　版本：原始四张分镜　历史记录：是
来源文件：提示词与参考记录.json
关联分镜：无对应公开分镜（早期版本）

生成一张16:9横向独立电影分镜静帧，photorealistic 真人电影质感，无文字、无边框、不是拼图。镜头03「沉默凝视」，中近景裁切至胸口下方，画面只出现一位原创成年男子。参考输入依次为：第1张仅负责参考视频中男主向左凝视的反打构图与含蓄表演，绝不复制其演员面孔、寸头、立领或民国服装；第2张仅负责白石宽灰脉、黑木和银灰墨染衣料材质，不继承其中两人的身份，不加入女人。现在首次设计全新的固定男主。
原创男主34岁，身材修长挺拔，脸型略长而下颌偏方，颧颊自然有体积；浓直眉、横长深褐眼睛、平直眼尾，挺直匀称的鼻梁与自然鼻尖，清楚唇线且下唇略厚，无胡须，眼周保留轻微成熟纹理，俊朗沉静而不刀削。黑发全部束成整洁高髻，窄黑灰布带固定，无冠冕。穿深灰内领、银灰交领墨染长袍、灰黑腰带。肩袖有大片浓淡墨染顺布面褶皱转折，细织衣料而不是粗麻。
与女主对坐的反打：男子胸肩位于画面右中部，身体和脸朝画面左方，左向三分之二侧面。眼睛清楚望向左侧画外坐着的女主，嘴自然闭合，呼吸平静，带有被问到感情后的克制与迟疑，不愤怒、不微笑、不看镜头。面部占画高约四成，头顶高髻完整且有余量，左侧留视线空间。双手在画面下缘外，不放刀剑；85mm级中长焦自然透视，与女主镜头保持眼平机位及相近景别。
背景是同一古典庭院廊下，洁净亮白石墙有连续宽幅灰矿脉，少量哑光黑木柱在右边缘框景，阴天庭院大面积漫射光从左前方塑脸，鼻侧和颧颊有连续柔转折、眼窝有填充。白石保留纹理、黑木保留暗层次，皮肤留极淡自然暖色；稍浅景深，宽灰石脉依然可辨。只改变反打机位，保持同一白壁无声材料系统与清净空间。避免复制第1张演员、寸头或民国立领；避免年轻偶像脸、尖下巴、整幅粗颗粒、塑料皮肤、纯黑背景和漂浮墨烟。


================================================================
prompt-003｜女主抬眼追问
类型：生图　版本：原始四张分镜　历史记录：否
来源文件：提示词与参考记录.json
关联分镜：storyboard-02

以实际输入的第1张已生成女主镜头01为画面基础，生成镜头02「抬眼追问」的一张16:9横向独立电影分镜静帧，photorealistic真人电影质感。第2张是用户指定的女主原始身份图，用于严格保持同一人五官与造型。只改变同一女主的头部角度、眼神与嘴部表情，其余维持镜头01：同一女主、同一镜头焦段、同一胸口以上中近景、同一机位高度、同一黑木柱和白石宽灰矿脉背景、同一白内领与银灰墨染交领外衣、同一低髻灰簪与鬓发、同一右前方庭院柔光。画面仍然只有一位女子，胸肩位于左中，身体朝右，右侧视线留白。
她将头轻轻抬起，视线从右下方抬至画面右侧画外男主的眼睛高度，仍是右向三分之二侧脸。眉心只有极轻的紧张，眼睛略睁开、湿润但没有眼泪，唇轻启，像认真追问感情所在的一瞬间；清丽而克制，带有期待与受伤，不做夸张哭戏、不看镜头。面容严格延续输入的原始女子：柔和椭圆轮廓、自然略长杏眼、细直鼻梁、清楚淡玫瑰唇形、小面积淡暖肤色与自然微小肤理。自然颧颊和眼睑体积，皮肤不磨成塑料。让衣上的大块浓淡墨染继续依附同一织物褶面，白石灰脉的排列与第1张保持一致。无文字、无边框、无拼图；不换人物、不加男子、不加茶杯或伞、不更改服装发簪、不把表情改成媚笑。


================================================================
prompt-004｜男主沉默凝视（构图修正版）
类型：生图　版本：原始四张分镜　历史记录：否
来源文件：提示词与参考记录.json
关联分镜：storyboard-03

对实际输入这张男主分镜做一项精准构图修正，仍输出16:9横向单张photorealistic真人电影静帧：保持同一男主的面部身份、眉眼鼻唇下颌、年龄、向左凝视、闭唇沉默、全部服装墨染纹样、白石灰脉背景、右侧黑木及左前方柔光。将整体机位景别略拉远约8%，取景轻轻向上调整，使完整的黑发高髻及发束顶部都进入画面，高髻最高点与画面上缘之间留出约画面高度5%的白石背景，不裁切头顶。人物仍位于右中部，左侧有视线空间，胸口以上中近景。保持眼神的迟疑和克制，眉心略放松但不改变人脸。除必要的画幅扩展和头顶留白外不改变画面，不加入任何人、物、文字、边框或胶片颗粒。


================================================================
prompt-005｜男主垂眸作答
类型：生图　版本：原始四张分镜　历史记录：否
来源文件：提示词与参考记录.json
关联分镜：storyboard-04

以实际输入这张已经修正头顶留白的男主镜头03为基础，生成镜头04「垂眸作答」：一张16:9横向独立photorealistic真人电影分镜静帧。仅改变这个同一成年男子的头部俯仰、视线和嘴部瞬间，其他维持输入：同一面孔、相同眉眼鼻唇下颌与34岁年龄、黑发整齐高髻和窄黑灰布带、深灰内领、银灰交领衣料及肩袖大片墨染、同一背景白石宽灰矿脉排列与右边缘黑木柱、同一机位焦段、人物右中部胸口以上裁切和左侧视线留白、同一左前方庭院宽幅漫射光与微暖肤色。高髻完整入画、最高点与画面上缘有约5%高度余量。
剧情瞬间：男子刚从左侧画外女主身上移开目光，头低下约15度，眼睛自然垂向左下方近处桌面，眉心紧张略缓，嘴唇极轻打开，像终于用很短的词作答的第一瞬；克制而沉重，带着欲言又止，没有笑、没有怒、没有流泪，不能闭成双眼睡着。他胸肩朝左，姿态稳，双手继续在画面下缘之外。鼻侧和下眼睑保持柔和连续明暗，脸不能因低头变成另一人或失去可读性。白石仍亮而有宽灰石脉，黑木仍有层次，衣染贴附织物而不漂浮；不要文字、边框、拼图，不加入女子、伞、刀剑、茶具或新景物，不让反打轴线改变。


================================================================
prompt-006｜01_女主轻声疑惑
类型：生图　版本：全国覆盖（原资料版本）　历史记录：否
来源文件：出图提示词与记录.json
关联分镜：storyboard-05

以输入图片为严格的同一位女主、同一套衣服、同一机位和同一庭院参考，制作电影连续分镜第一张。16:9横画幅，单张完整真人电影静帧，无黑边，无拼图，无字。胸部以上近景，女子位于画面左侧，向右看画外同高度坐着的男主，右侧留视线空间。保持参考脸型、眉眼鼻唇比例、低盘黑色编发髻、灰色素簪和两绺自然鬓发；保持原图银灰层叠交领、大片墨染随衣褶的细织衣料。当前瞬间是听到男主回答后轻声疑惑地反问的起点：上眼睑略抬，眉心轻微收拢，眼里仍有认真与一点委屈，双唇自然微分，嘴角不笑，头和肩端正，身体尚未前倾。不要瞪眼和大张嘴。左侧黑木立柱与背后带连续宽灰色矿脉的亮白石墙保持原来的构图与纹样位置。柔和阴天光从右前方塑造自然面孔，细腻真实皮肤有淡暖色，背景衣料近乎黑白灰，白石保留层次。不增添人物、道具、现代元素、字幕、品牌文字或水印；保留发髻轮廓，头顶留适当空间，手在画外。干净自然成像，无粗胶片颗粒、纸纹或塑料磨皮。输入图是角色身份、衣装与空间参考，不是要生成分屏对比。


================================================================
prompt-007｜02_女主突然爆发
类型：生图　版本：全国覆盖（原资料版本）　历史记录：否
来源文件：出图提示词与记录.json
关联分镜：storyboard-06

在所给电影分镜参考图上进行连续性表演变化：生成同一位女主突然提高声音质问的第二个表演状态。16:9，独立单张真人电影静帧，不是拼图。保持输入图的相机位置、焦距、胸部以上裁切、人物偏左的位置、向画右男主的视线，固定背景、低盘编发髻、素灰发簪、少量鬓发和银灰交领墨染长衣，保持同一张自然面孔。只明显改变神态和说话口型：女子眉心收紧，双眼有明确质问的力量，眼睛正常大小，头仍端正，肩微绷，正在突然提高嗓音的一瞬间，自然张开嘴说话，下颌适度下移，唇形清楚，少量上牙自然可见，不露出夸张牙龈。她是真生气，不带笑，不做鬼脸，不哭。这个瞬间的爆发须比参考的柔和疑惑明显升级。仍坐着，身体尚未大幅前倾，双手留在画外，不指人。亮白石墙宽幅连续灰脉、左边黑木柱、干净灰石地保持原位置，柔光从前右侧塑脸，衣料大块墨染附着真实褶面，皮肤淡暖且保留真实细微纹理。画面不出现任何文字、字幕、logo、酒店招牌、手机或现代道具，不生成额外人物、漫画夸张表情、粗颗粒或磨皮脸。图用于同一连续镜头中的中间表演关键帧，不能变成新机位或更近的特写。


================================================================
prompt-008｜03_女主前倾追问
类型：生图　版本：全国覆盖（原资料版本）　历史记录：否
来源文件：出图提示词与记录.json
关联分镜：storyboard-07

制作这段真人电影的第三个连续表演关键帧：同一女子越说越气，小幅前倾继续追问。图1负责同一位女主身份、原始胸上构图、衣装、发髻发簪和固定庭院背景；图2负责已经升级的怒意与说话状态。生成独立16:9横幅完整静帧，无拼图文字。相机必须固定在原位置，不推近不换角度：女子仍在左侧，面对画面右边的画外男主，只是坐姿从腰背自然前倾约8度，肩胸及头一同向右前方小幅移动，她的脸因姿势变化略向右下移动但仍保留右侧视线空间，别变成巨大头部特写。眉心比图2更收紧，眼神锁定对方，有认真发火的压迫感，眼球不夸张睁大；说话的嘴自然适度张开，下颌正常，嘴角不扬，不哭不搞怪，避免夸张露齿。此图是较大声说一句追问的中间瞬间，不是永远固定张嘴。双手保持画外，不起身不指人不拍桌。延续同样椭圆脸、自然杏眼与鼻唇比例、低盘黑色编发髻、灰色素簪、零散鬓发、银灰层叠交领和大片墨染细织衣料；衣褶随小幅前倾自然变化，不能换衣服。参考中左侧黑木立柱、背后连续宽幅灰色矿脉的亮白石墙与灰石地始终不动；右前方宽幅柔光，淡暖自然肤色，清楚连续面部体积，无塑料磨皮无粗颗粒。单人、无字幕、无可见文字、无品牌标志、无现代设备、无酒店建筑，喜感只来自动作与一本正经的愤怒，不来自漫画化造型。


================================================================
prompt-009｜04_男主抬眼语塞
类型：生图　版本：全国覆盖（原资料版本）　历史记录：否
来源文件：出图提示词与记录.json
关联分镜：storyboard-08

以输入图片为严格的同一名男主身份、发髻、衣装、机位和庭院空间参考，生成古装真人电影的结尾反应分镜：男子被画外女子突然大声质问后抬眼语塞的瞬间。独立16:9横幅单张，无边框无拼图无文字。保持男子坐在画面右侧、胸部以上近景、朝向画左的空间关系、完整黑色高髻与束带、深灰内领和银灰大片墨染交领长衣。相机保持不动，同样焦距和裁切，头頂以上留空间。与输入低头图相比只改变头眼和很小的表情：下巴略抬，视线从左下抬到画左与对方眼睛相对的高度，眉梢轻抬、眼睛略微睁开但仍是正常眼形，嘴唇微分一点像想说话又没说出来，既尴尬又被问住，保持成熟沉静的真实人脸。不要笑，不皱成怒脸，不做惊恐漫画表情，不张大嘴，不后仰闪躲，不耸肩。脸型、浓直眉、横长眼、鼻梁、下颌和成熟自然皮肤纹理保持同一个人。保持输入白石墙连续宽灰矿脉、右边深色木栏、远处模糊松形和衣料纹样，左前方宽幅柔光与淡暖肤色一致。眼睛和唇部有柔和清楚体积，白石亮而有层次。双手在画外，单人，无字幕、品牌文字、logo、手机、现代物品，无胶片粗粒、塑料磨皮、水墨插画效果。此图只表现抬眼语塞一瞬，后续再低头由视频提示词表达，不要在单张里叠加多个头或多重曝光。


================================================================
prompt-010｜装委屈辩解
类型：生图　版本：两省一市_男主结尾　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-09

制作一张独立完整的16:9横向真人电影分镜参考图，目标2048×1152。只出现一位成年男子，胸部以上近景，沿用输入图1的固定机位、人物右侧构图、朝向画外左边的视线和左侧留白，完整保留头顶发髻。输入图1负责男主身份、服装、场景及构图，输入图2辅助保持同一个人面部与说话表情，不借用图2的惊慌情绪。必须保持这两张图中的同一个成熟男子：浓直眉、横长眼睛、挺直匀称鼻梁、清楚连续的颧颌轮廓、自然唇形、淡暖肤色和成熟皮肤细纹，干净无胡须，黑发整齐束高髻并系深灰布带。银灰层叠交领衣袍，大块浓淡墨染顺着肩胸真实衣褶下垂。背景是同一庭院明亮白石墙，宽幅连续灰色矿石纹向斜侧延伸，画面极右保持深色木柱与竖木格。庭院左前方宽幅自然柔光，脸部有连续柔和体积，环境白灰黑，肤色微暖。以下是同一段喜剧表演中的一个瞬间，改变表情而非换人换场景。他正在故作无辜地认真辩解：头略低、下巴微收，眉毛内侧轻轻抬起，眼睛直视画外左侧的女主，嘴唇轻微分开处于自然说话状态，肩颈端正，脸上暂时没有笑容。有一点假委屈和装正经的意味，不能真哭或愤怒。画面不出现任何文字、对白、字幕、编号、边框、水印、拼图或分屏；不出现女主、不新增其他人物。保持原人物骨相和衣纹，不把成熟男主变成年轻偶像，不磨皮、不加胶片粗颗粒，不画成水墨插画。


================================================================
prompt-011｜抬眼露笑
类型：生图　版本：两省一市_男主结尾　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-10

制作一张独立完整的16:9横向真人电影分镜参考图，目标2048×1152。只出现一位成年男子，胸部以上近景，沿用输入图1的固定机位、人物右侧构图、朝向画外左边的视线和左侧留白，完整保留头顶发髻。输入图1负责男主身份、服装、场景及构图，输入图2辅助保持同一个人面部与说话表情，不借用图2的惊慌情绪。必须保持这两张图中的同一个成熟男子：浓直眉、横长眼睛、挺直匀称鼻梁、清楚连续的颧颌轮廓、自然唇形、淡暖肤色和成熟皮肤细纹，干净无胡须，黑发整齐束高髻并系深灰布带。银灰层叠交领衣袍，大块浓淡墨染顺着肩胸真实衣褶下垂。背景是同一庭院明亮白石墙，宽幅连续灰色矿石纹向斜侧延伸，画面极右保持深色木柱与竖木格。庭院左前方宽幅自然柔光，脸部有连续柔和体积，环境白灰黑，肤色微暖。以下是同一段喜剧表演中的一个瞬间，改变表情而非换人换场景。他的假委屈开始露馅：下巴比上一刻略微抬起，眉心舒展开，一侧眉毛轻轻上扬，目光仍朝画外左边的女主，嘴唇刚说完一句话自然合拢，一侧嘴角已经抑制不住向上提，出现清晰可读的轻微得意坏笑。表情像在一本正经纠正对方统计错误，觉得自己很有道理。画面不出现任何文字、对白、字幕、编号、边框、水印、拼图或分屏；不出现女主、不新增其他人物。保持原人物骨相和衣纹，不把成熟男主变成年轻偶像，不磨皮、不加胶片粗颗粒，不画成水墨插画。


================================================================
prompt-012｜得意报地名
类型：生图　版本：两省一市_男主结尾　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-11

制作一张独立完整的16:9横向真人电影分镜参考图，目标2048×1152。只出现一位成年男子，胸部以上近景，沿用输入图1的固定机位、人物右侧构图、朝向画外左边的视线和左侧留白，完整保留头顶发髻。输入图1负责男主身份、服装、场景及构图，输入图2辅助保持同一个人面部与说话表情，不借用图2的惊慌情绪。必须保持这两张图中的同一个成熟男子：浓直眉、横长眼睛、挺直匀称鼻梁、清楚连续的颧颌轮廓、自然唇形、淡暖肤色和成熟皮肤细纹，干净无胡须，黑发整齐束高髻并系深灰布带。银灰层叠交领衣袍，大块浓淡墨染顺着肩胸真实衣褶下垂。背景是同一庭院明亮白石墙，宽幅连续灰色矿石纹向斜侧延伸，画面极右保持深色木柱与竖木格。庭院左前方宽幅自然柔光，脸部有连续柔和体积，环境白灰黑，肤色微暖。以下是同一段喜剧表演中的一个瞬间，改变表情而非换人换场景。他说话时越来越得意：保持朝画外左侧女主的目光，下巴自然抬高一点，脸轻轻倾斜，眉眼带狡黠的自信，嘴唇自然分开正顺畅报出一串地名，嘴角带着明显笑意，面颊略抬，眼睛仍然睁开可读。他是在炫耀和自曝，身体放松；不要增加抬手、数手指或道具，不要变成仰头大笑。画面不出现任何文字、对白、字幕、编号、边框、水印、拼图或分屏；不出现女主、不新增其他人物。保持原人物骨相和衣纹，不把成熟男主变成年轻偶像，不磨皮、不加胶片粗颗粒，不画成水墨插画。


================================================================
prompt-013｜收口得意偷笑
类型：生图　版本：两省一市_男主结尾　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-12

制作一张独立完整的16:9横向真人电影分镜参考图，目标2048×1152。只出现一位成年男子，胸部以上近景，沿用输入图1的固定机位、人物右侧构图、朝向画外左边的视线和左侧留白，完整保留头顶发髻。输入图1负责男主身份、服装、场景及构图，输入图2辅助保持同一个人面部与说话表情，不借用图2的惊慌情绪。必须保持这两张图中的同一个成熟男子：浓直眉、横长眼睛、挺直匀称鼻梁、清楚连续的颧颌轮廓、自然唇形、淡暖肤色和成熟皮肤细纹，干净无胡须，黑发整齐束高髻并系深灰布带。银灰层叠交领衣袍，大块浓淡墨染顺着肩胸真实衣褶下垂。背景是同一庭院明亮白石墙，宽幅连续灰色矿石纹向斜侧延伸，画面极右保持深色木柱与竖木格。庭院左前方宽幅自然柔光，脸部有连续柔和体积，环境白灰黑，肤色微暖。以下是同一段喜剧表演中的一个瞬间，改变表情而非换人换场景。台词全部说完后的结尾表情：头略微向下点，眼睛仍向左侧女主斜瞥，眼角因笑意自然收窄，面颊上提，嘴唇合拢、两边嘴角忍不住向上扬，其中一边更高，形成明确有戏的狡黠得意偷笑。肩膀放松，像自以为解释得非常完美，露馅却完全没有意识到。笑意比前面明显，幅度自然真实，没有咧嘴露大排牙齿，没有邪恶怪脸，不看镜头。画面不出现任何文字、对白、字幕、编号、边框、水印、拼图或分屏；不出现女主、不新增其他人物。保持原人物骨相和衣纹，不把成熟男主变成年轻偶像，不磨皮、不加胶片粗颗粒，不画成水墨插画。


================================================================
prompt-014｜01_女主称呼与发问_8秒_I2VA
类型：视频　版本：H3提示词 / I2VA　历史记录：否
来源文件：01_女主称呼与发问_8秒_I2VA.txt
关联分镜：storyboard-01

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene, 16:9 framing inherited from <Picture 1>. Begin from its exact chest-up composition, pictured female identity, low black bun, plain gray hairpin, white inner collar, ink-dyed silver-gray robe, screen-left position, lowered rightward gaze, dark timber at frame-left and white stone with broad gray veins. Soft front-right courtyard light preserves faint warm skin color. The camera holds a static shot throughout; hands stay outside the frame.
At approximately 00:00.000-00:01.500, the woman (S1), using a soft adult female Mandarin voice, makes small natural lip movements and says: <d>[Chinese] 汉卿啊.</d> Her head stays slightly inclined. During approximately 00:01.500-00:05.750, preserve the pause: eyes remain directed down-right, mouth corners settle, lips gently gather and may part a little with quiet breathing. No nodding, dramatic shoulder movement or hand gesture.
At approximately 00:05.750, the same woman (S1) asks: <d>[Chinese] 你的感情在哪儿啊?</d> Her gaze starts low. Around 00:06.200, her eyes rise toward the off-screen man to screen-right and her chin follows with a small lift; by about 00:06.500, she holds eye contact. Her upper eyelids open slightly, the inner brows show restrained tension and the lips articulate the question without exaggeration. Near 00:07.500, the last syllable settles and her lips remain lightly parted. Hold that raised-gaze state until 00:08.000. Keep the same face, hairpin, loose strands and ink-dyed fabric pattern; no cuts, tears, smile, turn away or new dialogue. No subtitles, captions, titles, logos or any on-screen text.

overall_soundscape: A low courtyard air tone and faint natural breathing, with only subtle fabric sound. No extra human voices or dramatic sound effects.

non_diegetic_music: N/A



================================================================
prompt-015｜02_男主沉默凝视_6秒_I2VA
类型：视频　版本：H3提示词 / I2VA　历史记录：否
来源文件：02_男主沉默凝视_6秒_I2VA.txt
关联分镜：storyboard-03

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene, framing inherited from <Picture 1>. Start from its exact chest-up male reverse shot: the pictured mature man occupies screen-right and faces screen-left, with his gaze already resting on the off-screen woman. Preserve his face, dark brows, eyes, natural nose and jaw, clean-shaven skin, complete black topknot, narrow dark band, dark-gray inner collar and silver-gray ink-dyed robe. Retain the white stone's broad gray veins, right-edge dark timber and soft front-left light. The camera remains static at eye level.
From 00:00.000 to approximately 00:01.500, hold his upward-left eye line, natural closed lips and almost motionless head. His shoulders stay still; do not insert a preparatory nod or turn. Between approximately 00:01.500 and 00:03.000, a brief, small natural eyelid movement may occur and the eyes return to the same off-screen target. This is optional actor direction, not a repeated blink pattern. His jaw stays relaxed but composed, mouth closed, with no speaking or smile.
For the rest of the six-second shot, let slight facial tension remain in the brows and mouth corners while maintaining the same gaze and posture. Do not look down yet: the downward state belongs to another shot. Hands remain outside the crop, costume markings stay fixed to the fabric and skin keeps a faint warm tone. Do not add dialogue, voiceover, nods, tears, gestures or camera movement. No subtitles, captions, titles, logos or any on-screen text.

overall_soundscape: A low courtyard air tone and faint natural breathing, with only subtle fabric sound. No extra human voices or dramatic sound effects.

non_diegetic_music: N/A



================================================================
prompt-016｜03_女主等待回应_4秒_I2VA
类型：视频　版本：H3提示词 / I2VA　历史记录：否
来源文件：03_女主等待回应_4秒_I2VA.txt
关联分镜：storyboard-02

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene. Begin at the exact chest-up composition and raised-gaze state of <Picture 1>. The pictured woman occupies screen-left, faces screen-right and is already looking at the off-screen man. Preserve her original face, low black bun, gray hairpin, loose strands, white inner collar, silver-gray ink-dyed robe and the same white stone veins, left-edge timber and soft front-right light. Her lips are lightly parted and the inner brows carry very small tension.
During 00:00.000-00:01.000, she holds her eye line rather than beginning another head lift. During 00:01.000-00:02.200, allow minute eyelid and lower-jaw adjustments without rhythmic talking-shaped mouth movement. The head, neck and shoulders remain almost still. She is waiting for an answer, and says nothing. From 00:02.200 to 00:04.000, extend the same silent pose for an editing tail, without introducing a new emotional event.
The camera remains static, hands stay outside the crop and no person enters the frame. Keep a consistent face, hairpin, robe pattern, wall pattern, exposure and faint warm skin color. No repeated question, additional speech, emphatic nod, turn away, downward gaze, smile or tears. No subtitles, captions, titles, logos or any on-screen text.

overall_soundscape: A low courtyard air tone and faint natural breathing, with only subtle fabric sound. No extra human voices or dramatic sound effects.

non_diegetic_music: N/A



================================================================
prompt-017｜04_男主低头作答_4秒_I2VA
类型：视频　版本：H3提示词 / I2VA　历史记录：否
来源文件：04_男主低头作答_4秒_I2VA.txt
关联分镜：storyboard-04

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene. Begin from the exact chest-up composition of <Picture 1>: the pictured mature man occupies screen-right, faces screen-left and is already looking down-left with his chin lowered. Preserve his face, complete black topknot and dark band, dark-gray inner collar, silver-gray ink-dyed robe, broad-veined white stone, right-edge timber and soft front-left light. Match the pictured opening mouth state. The camera stays static.
During approximately 00:00.000-00:00.550, he maintains the low chin and downward gaze, with no preceding eye contact or visible new head-lowering movement. Between approximately 00:00.550 and 00:01.100, the man (S1), using a low, measured adult male Mandarin voice, gives one short reply: <d>[Chinese] 家.</d> The lips and jaw move only as much as the single syllable requires; do not append another word or repeat it. By approximately 00:01.450, his lips return toward a light closure. Maintain the same lowered head and quiet, composed expression until 00:01.700.
From 00:01.700 to 00:04.000, extend this silent downward-gaze state as an editing tail. Do not add a smile, grimace, nod, look up, new line or hand gesture. Keep identity, eye shape, nose and jaw, hairstyle, fabric markings and white stone veins consistent; shadows under the brow remain soft enough to read the eyes. No subtitles, captions, titles, logos or any on-screen text.

overall_soundscape: A low courtyard air tone and faint natural breathing, with only subtle fabric sound. No extra human voices or dramatic sound effects.

non_diegetic_music: N/A



================================================================
prompt-018｜05_女主听后停住_4秒_I2VA
类型：视频　版本：H3提示词 / I2VA　历史记录：否
来源文件：05_女主听后停住_4秒_I2VA.txt
关联分镜：storyboard-02

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene. Begin from the exact pictured chest-up female composition and existing raised-gaze pose of <Picture 1>. The same woman sits on the left and faces the off-screen man to screen-right, with lightly parted lips and attentive eyes. Preserve her face, low bun, plain gray hairpin, loose strands, white inner collar and silver-gray ink-dyed robe. Retain the same broad gray veins on the white stone wall, left-edge dark timber, soft front-right light and faint warm skin color. The camera holds a static shot throughout.
During approximately 00:00.000-00:01.300, she remains silent and receives the answer without a sudden facial reaction. The inner brows and lip tension may change by a very small amount; her eye line stays on the man. During approximately 00:01.300-00:03.034, let that nearly motionless reaction continue: lips still lightly separated, chin and shoulders stable, eyes neither turning away nor dropping. No additional question or continuous speaking-shaped mouth motion. The scene's emotional weight comes from holding the gaze.
From approximately 00:03.034 to 00:04.000, maintain the same quiet state as an editing tail. Do not add crying, a falling tear, a smile, a pronounced mouth closure, a nod, a head turn, a hand movement or an off-screen line. Keep identity, hairpin, fabric pattern, wall veins and lighting continuous. No subtitles, captions, titles, logos or any on-screen text.

overall_soundscape: A low courtyard air tone and faint natural breathing, with only subtle fabric sound. No extra human voices or dramatic sound effects.

non_diegetic_music: N/A



================================================================
prompt-019｜A_前段_14秒_H3官方结构
类型：视频　版本：H3提示词 / Ref2VA　历史记录：否
来源文件：A_前段_14秒_H3官方结构.txt
关联分镜：storyboard-01、storyboard-02、storyboard-03

subject_definitions:
<Subject 1> is the adult woman in <Picture 1> and <Picture 2>: the same oval face, long almond-shaped eyes, straight slender nose, pale rose lips, low black braided bun, plain gray hairpin, white inner collar and silver-gray ink-dyed cross-collar robe. Her performance follows the woman's small facial changes in <Video 1>, while her identity and clothing come only from the pictures.
<Subject 2> is the adult man in <Picture 3>: a mature, slightly long face, straight dark brows, long eyes, balanced nose and jaw, no beard, black topknot with a narrow dark band, dark-gray inner collar and silver-gray ink-dyed robe.
<Subject 3> is the courtyard setting shared by the pictures: clean white stone with broad continuous gray mineral veins, dark timber posts and soft overcast light.
<Picture 1> is the lowered-gaze composition anchor for [Shot 1].
<Picture 2> is the raised-gaze performance anchor reached later within [Shot 1], without introducing a cut.
<Picture 3> is the male reverse-shot composition anchor for [Shot 2].
<Video 1> is the 13.033-second opening source excerpt. It supplies restrained acting and the cut at 00:07.733, not the source actors, costumes, colored interior or burned-in captions.
<Audio 1> is the matching source-excerpt audio, used as a reference for the woman's dialogue content and measured delivery, associated with <Subject 1> (S1); it is not copied as the final soundtrack.

summary:
[reference generation + audio reference] Create a 14-second, 16:9 live-action exchange in <Subject 3>. <Subject 1> addresses the off-screen man, pauses, then lifts her eyes while asking about his feelings. At 00:07.733, cut to <Subject 2> listening without speaking. Preserve the source performance rhythm through 00:13.033, then extend the man's silent hold to 00:14.000.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the pictured identity, low bun, hairpin, collar and ink-dyed robe throughout her facial changes.
<Subject 2> (appears in [Shot 2]): fully_preserved - retain the pictured face, topknot and clothing.
<Subject 3> (appears in [Shot 1], [Shot 2]): fully_preserved - retain white stone veins, timber geometry and directional soft light.
<Picture 1> ([Shot 1] opening composition): partially_preserved - retain framing and lowered gaze while lips and eyelids move naturally.
<Picture 2> ([Shot 1] raised-gaze anchor): partially_preserved - reach the pictured eye line and head angle during the question, not at a new cut.
<Picture 3> ([Shot 2] composition): partially_preserved - retain framing, leftward gaze and complete topknot while allowing restrained facial motion.
<Video 1> (acting and cut structure): partially_preserved - transfer its facial performance and cut timing; replace people and setting, omit all captions, and add only the final silent hold.
<Audio 1>: reference - reperform the same Chinese words with restrained conversational delivery; do not copy the original signal.

detailed_description:
The target is a restrained live-action period scene. White, gray and black come from stone, timber and fabric; skin keeps a faint natural warmth. No subtitles, captions, titles, logos or other on-screen text appear at any time.
[Shot 1] <Subject 1> sits in a chest-up medium close-up on the left side, facing the off-screen man to screen-right. The opening follows <Picture 1>: her chin is slightly lowered, eyes directed down-right, shoulders still and hands outside the crop. The dark post stays at frame-left; broad gray veins remain visible on the white wall behind her. Soft courtyard light from front-right models her eyelids, nose and cheeks. The camera holds a static shot, with no push, pan, reframing or focus pull.
During approximately 00:00.000-00:01.500, <Subject 1> (S1), speaking in a soft adult female Mandarin voice, addresses him with a small, natural opening and closing of her lips: <d>[Chinese] 汉卿啊.</d> Her gaze remains lowered and the phrase is directed to him rather than to the camera. Between approximately 00:01.500 and 00:05.750, preserve the long pause: her lips gather gently, the mouth corners settle, and her head stays slightly inclined. Allow only tiny facial adjustments; no nodding, new hand gesture or emphatic shoulder movement.
At approximately 00:05.750, <Subject 1> (S1) asks, following the conversational phrasing referenced by <Audio 1>: <d>[Chinese] 你的感情在哪儿啊?</d> Begin with the gaze still low. Around 00:06.200, her eyes lift toward the man and the chin follows by a small amount; by approximately 00:06.500, she holds the raised eye line of <Picture 2>. The upper eyelids open slightly, the inner brows become subtly tense and the lips articulate the question without an exaggerated open mouth. Let the final syllable settle, then maintain eye contact with lips lightly parted. Keep her hairpin, loose strands and robe pattern stable.
[Shot 2] At 00:07.733, the camera cuts directly to <Subject 2> in the reverse composition of <Picture 3>. His chest and head occupy frame-right; his face and eyes are turned toward screen-left, with gaze space on the left. The white veined wall and timber at the right edge belong to the same courtyard. Soft light now reaches his face from front-left. He is already looking at the woman when the cut occurs. His lips remain closed, brows composed, head and shoulders nearly still. A brief natural eyelid movement is acceptable, but do not impose repeated blinking or a fixed blink count. He neither answers, nods, smiles nor looks down in this shot. Hold the unresolved gaze through 00:13.033, then continue the same quiet state until 00:14.000.

overall_soundscape:
A very low courtyard air tone and faint natural breathing sit under the dialogue. Keep incidental fabric sound subtle. The source audio supplies speech-reference information only; add no footsteps, dramatic effects or off-screen voices.

non_diegetic_music:
N/A



================================================================
prompt-020｜B_后段_7秒_H3官方结构
类型：视频　版本：H3提示词 / Ref2VA　历史记录：否
来源文件：B_后段_7秒_H3官方结构.txt
关联分镜：storyboard-02、storyboard-04

subject_definitions:
<Subject 1> is the adult woman in <Picture 1>: the same oval face, long almond-shaped eyes, pale rose lips, low black braided bun, plain gray hairpin and silver-gray ink-dyed cross-collar robe.
<Subject 2> is the adult man in <Picture 2>: the same mature face, straight dark brows, long eyes, natural jaw, clean-shaven skin, black topknot, narrow dark band, dark-gray inner collar and silver-gray ink-dyed robe.
<Subject 3> is the courtyard setting shared by the pictures: luminous white stone with broad gray mineral veins, dark timber geometry and soft directional overcast light.
<Picture 1> is the woman's raised-gaze composition reference for [Shot 1] and [Shot 3], not an instruction for her to repeat a question.
<Picture 2> is the man's already-lowered head and downward gaze reference for [Shot 2].
<Video 1> is the 6.934-second source excerpt beginning at original time 00:13.033. Its local cuts at 00:02.200 and 00:03.900 supply the acting and edit rhythm; its original people, colored interior and captions are excluded.
<Audio 1> is the matching source-excerpt audio, used only to guide the male one-word reply and restrained delivery of <Subject 2> (S1); the source signal is not copied.

summary:
[reference generation + audio reference] Create a 7-second, 16:9 live-action continuation in <Subject 3>. <Subject 1> silently waits with her gaze already raised. At 00:02.200, cut to <Subject 2>, who is already looking down and gives the brief reply. At 00:03.900, return to <Subject 1> holding a quiet reaction. No subtitles or additional dialogue.

retention_analysis:
<Subject 1> (appears in [Shot 1], [Shot 3]): fully_preserved - retain identity, low bun, hairpin, collar and robe across both reactions.
<Subject 2> (appears in [Shot 2]): fully_preserved - retain identity, topknot and ink-dyed robe.
<Subject 3> (appears in [Shot 1], [Shot 2], [Shot 3]): fully_preserved - retain the white stone, timber, light direction and tonal continuity.
<Picture 1> ([Shot 1], [Shot 3] composition): partially_preserved - retain position and eye line while permitting small changes in eyelids and lips.
<Picture 2> ([Shot 2] composition): partially_preserved - retain the already-lowered head and left-down gaze while articulating one word.
<Video 1> (acting and local cut structure): partially_preserved - transfer the two cut times and restrained performance; replace the source identities and setting, remove captions and extend only the final hold by about 0.066 seconds.
<Audio 1>: reference - reperform the source reply in Chinese without copying the waveform; use one male speaker and no new female line.

detailed_description:
The target continues the same clean, muted live-action period scene. Broad gray stone veins and large ink-dyed fabric areas remain attached to real surfaces; skin stays subtly warm. No subtitles, captions, text overlays, titles or logos appear anywhere.
[Shot 1] <Subject 1> is already seated in the raised-gaze chest-up composition of <Picture 1> at the opening. Her chest and face occupy the left side, facing screen-right toward the off-screen man. Keep the same low bun, hairpin, loose strands and cross-collar layers. A dark timber post stays at frame-left; the softly lit white veined wall occupies the background. Her chin is already lifted, eyes focused on him, lips slightly separated and inner brows subtly tense. Do not start a new upward head movement, repeat the previous question or introduce another line. Throughout 00:00.000-00:02.200, hold this waiting state: very small eyelid and jaw adjustments, steady shoulders, no nod, no tears and no hand entering the frame. The camera remains static at eye level, with no zoom or focus pull.
[Shot 2] At 00:02.200, the camera cuts to <Subject 2> in the composition and head position of <Picture 2>. He occupies the right side and faces screen-left. At the first frame of this shot he is already looking down-left, chin lowered; do not first show him making eye contact and then lowering his head. His upper body remains upright and restrained. Front-left soft light keeps the eyes, nose and cheek readable beneath the lowered brow, while the right-edge timber and white stone maintain the reverse-shot geography. Between approximately 00:02.700 and 00:03.250, <Subject 2> (S1), in a low, measured adult male Mandarin voice and following the restrained delivery referenced from <Audio 1>, gives a short reply with restrained lip and jaw movement: <d>[Chinese] 家.</d> The syllable is quiet and definite, without adding a breathy extra word or repeating it. He keeps his gaze down, allows the lips to return toward a light closure and holds his chin low until the cut. There is no smile, angry tightening, emphatic nod or gesture.
[Shot 3] At 00:03.900, the camera cuts directly back to <Subject 1> in the same screen-left, rightward-gaze composition used in [Shot 1]. She is silent. Her gaze remains on the man; lips stay lightly parted, with only a minute change in their tension as she receives the answer. Keep the head angle and shoulders steady. Do not invent crying, looking away, looking down, a second question or a pronounced mouth closure. Let the response remain visible through approximately 00:06.934, then sustain that same small, quiet state until 00:07.000. Retain consistent face shape, hairpin, clothing pattern, background veins and soft light across the two female shots.

overall_soundscape:
Low courtyard air tone and very faint natural breathing continue across the cuts. Keep the single reply intelligible and all other human voices absent.

non_diegetic_music:
N/A





================================================================
prompt-021｜05_女主反问_4秒_I2VA
类型：视频　版本：H3提示词 / 加台词版　历史记录：否
来源文件：05_女主反问_4秒_I2VA.txt
关联分镜：storyboard-02

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene, lasting 4 seconds. Begin from the exact chest-up female composition and raised-gaze pose of <Picture 1>. The same woman sits on the left, looking toward the off-screen man to screen-right. Preserve her face, low black braided bun, plain gray hairpin, loose strands, white inner collar and silver-gray ink-dyed robe. Retain the broad gray mineral veins on the luminous white stone wall, left-edge dark timber, soft front-right light and faint natural warmth in her skin. Keep the camera static throughout.
From 00:00.000 to approximately 00:00.450, she silently absorbs the answer given immediately before this clip. Her gaze stays fixed on the man, inner brows draw together slightly and lips gather gently; her chin and shoulders stay steady. At approximately 00:00.450-00:00.850, the woman (S1), in a soft adult female Mandarin voice with a clear mid-low register and controlled breath, repeats his word with a small questioning inflection: <d>[Chinese] 家?</d> She pauses for about 0.300 seconds, keeping the same eye line and allowing her jaw to settle. At approximately 00:01.150-00:02.400, the woman (S1), in the same quiet voice with restrained hurt and firmer emphasis on the first-person word, asks: <d>[Chinese] 那我算什么?</d> Synchronize natural compact lip articulation to each syllable. Let the final syllables settle quietly, without shouting, sobbing, a mocking smile or a head thrust.
From approximately 00:02.400 to 00:03.034, her lips relax toward a light closure and she holds her gaze on him. Continue the same silent state to 00:04.000 as an editing tail. The man stays off-screen and silent throughout this clip; do not repeat his earlier reply. Preserve face shape, hairpin, fabric pattern, wall veins and lighting. No tears, looking away, gestures, extra dialogue, subtitles, captions, titles, logos or any on-screen text.

overall_soundscape: A low courtyard air tone and faint natural breathing, with subtle fabric sound. Keep the pauses clear; add no other human voice or dramatic sound effects.

non_diegetic_music: N/A



================================================================
prompt-022｜B_后段_7秒_女主反问_Ref2VA
类型：视频　版本：H3提示词 / 加台词版　历史记录：否
来源文件：B_后段_7秒_女主反问_Ref2VA.txt
关联分镜：storyboard-02、storyboard-04

subject_definitions:
<Subject 1> is the adult woman in <Picture 1>: the same oval face, long almond-shaped eyes, pale rose lips, low black braided bun, plain gray hairpin and silver-gray ink-dyed cross-collar robe.
<Subject 2> is the adult man in <Picture 2>: the same mature face, straight dark brows, long eyes, natural jaw, clean-shaven skin, black topknot, narrow dark band, dark-gray inner collar and silver-gray ink-dyed robe.
<Subject 3> is the courtyard setting shared by the pictures: luminous white stone with broad gray mineral veins, dark timber geometry and soft directional overcast light.
<Picture 1> is the woman's raised-gaze composition reference for [Shot 1] and [Shot 3]; it anchors her silent waiting and subsequent spoken challenge.
<Picture 2> is the man's already-lowered head and downward gaze reference for [Shot 2].
<Video 1> is the source excerpt beginning at original time 00:13.033. Its local cuts at 00:02.200 and 00:03.900 guide the edit structure. The final female reaction is deliberately adapted into a new spoken challenge; source people, interior and captions are excluded.
<Audio 1> is the matching source-excerpt audio, used only to guide the male one-word reply and restrained delivery of <Subject 2> (S1). The waveform is not copied, and this audio does not supply or constrain the woman's new line.

summary:
[reference generation + audio reference] Create a 7-second, 16:9 live-action continuation in <Subject 3>. <Subject 1> silently waits with her gaze raised. At 00:02.200, cut to <Subject 2>, already looking down, for his brief reply. At 00:03.900, return to <Subject 1>, who pauses, repeats his word as a question, then delivers a quiet, pointed challenge while maintaining eye contact. Keep all dialogue audible with no subtitles.

retention_analysis:
<Subject 1> (appears in [Shot 1], [Shot 3]): fully_preserved - retain identity, low bun, hairpin, collar and robe.
<Subject 2> (appears in [Shot 2]): fully_preserved - retain identity, topknot and ink-dyed robe.
<Subject 3> (appears in all shots): fully_preserved - retain white stone, timber, light direction and tonal continuity.
<Picture 1> ([Shot 1], [Shot 3] composition): partially_preserved - retain framing and eye line, allowing natural speech articulation and restrained facial tension in the last shot.
<Picture 2> ([Shot 2] composition): partially_preserved - retain the lowered head and left-down gaze while articulating one word.
<Video 1> (local cut structure): partially_preserved - retain the two cut times and the opening waiting and reply beats; deliberately replace the final silent reaction with the woman's new line.
<Audio 1>: reference - reperform only the male source reply without copying the waveform. Generate the new female line independently; the source's final silence does not suppress her speech.

detailed_description:
The target continues the same clean, muted live-action period scene. Broad gray stone veins and large ink-dyed fabric areas remain attached to real surfaces; skin stays subtly warm and naturally textured. No subtitles, captions, text overlays, titles or logos appear anywhere.
[Shot 1] <Subject 1> begins in the raised-gaze chest-up composition of <Picture 1>. Her chest and face occupy the left side, facing screen-right toward the off-screen man. Preserve her low bun, hairpin, loose strands and layered cross-collar robe. A dark timber post remains at frame-left; the softly lit white veined wall fills the background. Her chin is already lifted, eyes focused on him, lips slightly separated and inner brows subtly tense. Throughout 00:00.000-00:02.200, she waits silently with small eyelid and jaw adjustments and steady shoulders. Do not repeat the earlier question. Keep the camera static at eye level, with no zoom or focus pull.
[Shot 2] At 00:02.200, the camera cuts to <Subject 2> in the composition and head position of <Picture 2>, following the local edit structure of <Video 1>. He occupies the right side, facing screen-left, and is already looking down-left with his chin lowered. His upper body stays upright. Front-left soft light keeps his eyes, nose and cheek readable; right-edge timber and white stone maintain the reverse-shot geography. Between approximately 00:02.700 and 00:03.250, <Subject 2> (S1), in a low, measured adult male Mandarin voice and following the restrained delivery referenced from <Audio 1>, replies with small natural lip and jaw movement: <d>[Chinese] 家.</d> He keeps his gaze down, lightly closes his lips after the word, and holds his chin low until the cut. He adds no further reply.
[Shot 3] At 00:03.900, the camera cuts directly back to <Subject 1> in the screen-left, rightward-gaze composition of <Picture 1>. Keep her gaze fixed on the off-screen man. From 00:03.900 to approximately 00:04.350, she absorbs his answer in silence: the inner brows draw together slightly and her lips gather gently. Her chin and shoulders remain steady. At approximately 00:04.350-00:04.750, <Subject 1> (S2), in a soft adult female Mandarin voice with a clear mid-low register and controlled breath, repeats his word with a small questioning inflection: <d>[Chinese] 家?</d> She pauses for about 0.300 seconds, keeping the same eye line and letting her jaw settle. At approximately 00:05.050-00:06.300, <Subject 1> (S2), in the same quiet voice with restrained hurt and firmer emphasis on the first-person word, asks: <d>[Chinese] 那我算什么?</d> Use clear, natural lip synchronization and compact articulation. The final syllables settle quietly, without shouting, sobbing, a mocking smile or a head thrust. By 00:06.300 her speech has finished; her lips relax toward a light closure and she continues looking at him through 00:07.000, leaving his silence unanswered. No looking away, tears, gestures, extra line or additional cut. Preserve her face shape, hairpin, robe pattern, background veins and soft light throughout. The camera remains static.

overall_soundscape:
Low courtyard air tone and faint natural breathing continue across the cuts, with subtle cloth sound. Keep the pauses clear and free of dramatic sound effects, crowd noise or additional voices.

non_diegetic_music:
N/A



================================================================
prompt-023｜B_后段15秒_全国覆盖_Ref2VA
类型：视频　版本：全国覆盖（原资料版本） / H3提示词　历史记录：否
来源文件：B_后段15秒_全国覆盖_Ref2VA.txt
关联分镜：storyboard-05、storyboard-06、storyboard-07、storyboard-04、storyboard-08

subject_definitions:
<Subject 1> is the same adult woman across <Picture 1>, <Picture 2> and <Picture 3>: oval face, almond-shaped eyes, pale rose lips, low black braided bun, gray hairpin, loose side strands and a silver-gray ink-dyed cross-collar robe.
<Subject 2> is the same adult man in <Picture 4> and <Picture 5>: mature face, straight brows, long eyes, clean-shaven jaw, black topknot and dark band, dark-gray inner collar and silver-gray ink-dyed robe.
<Subject 3> is the shared classical courtyard: luminous white stone with broad gray mineral veins, restrained dark timber, gray paving and soft directional overcast light.
<Picture 1> guides the woman's chest-up framing and quiet raised gaze in [Shot 1] and the opening of [Shot 3].
<Picture 2> guides her tightened brows and sudden vocal outburst within [Shot 3].
<Picture 3> guides her small forward lean and sustained questioning within the same [Shot 3]. These three pictures show one woman at successive performance stages, not three women or three cuts.
<Picture 4> guides the man's lowered head and down-left gaze in [Shot 2].
<Picture 5> guides the man's raised eyes and speechless reaction at the start of [Shot 4].

summary:
[reference generation] Create a 15-second, 16:9 live-action period-comedy scene in <Subject 3>. <Subject 1> waits, <Subject 2> quietly answers, and she shifts from soft disbelief to a sudden loud challenge and an indignant follow-up. End on his silent, embarrassed reaction. Use all five pictures for identity, framing and performance guidance. The humor comes from her sincerely angry delivery of modern brand references inside the serious classical setting. Generate the specified Chinese dialogue with synchronized lips; show no subtitles or modern objects.

retention_analysis:
<Subject 1> (appears in [Shot 1], [Shot 3]): fully_preserved - retain identity, bun, hairpin and robe throughout her changing performance.
<Subject 2> (appears in [Shot 2], [Shot 4]): fully_preserved - retain identity, topknot and robe across the reverse shots.
<Subject 3> (appears in all shots): fully_preserved - retain stone, timber, light direction and material color.
<Picture 1> ([Shot 1], [Shot 3] framing): partially_preserved - retain seated position and rightward eye line while animating speech.
<Picture 2> ([Shot 3] outburst): partially_preserved - transfer brow tension and forceful articulation, not a permanently open mouth.
<Picture 3> ([Shot 3] follow-up): partially_preserved - transfer the modest seated forward lean without changing camera position or identity.
<Picture 4> ([Shot 2] reply): partially_preserved - retain the lowered pose while articulating one word.
<Picture 5> ([Shot 4] reaction): partially_preserved - start with raised eyes and parted lips, then animate a silent retreat into a lowered gaze.

detailed_description:
Live-action cinematic realism, clean white-gray-black materials, faint warm skin and natural facial texture. Broad mineral veins remain attached to stone and ink-dyed areas follow real robe folds. The classical courtyard remains unchanged when modern brands are mentioned. No subtitles, captions, written dialogue, logos, signs, phones, hotel inserts or on-screen text.
[Shot 1] <Subject 1> sits on frame-left in the chest-up composition of <Picture 1>, facing the off-screen man to screen-right. Her chin is already raised, eyes focused on him, lips lightly separated. Preserve her low bun, gray hairpin, layered collar and shoulder ink patterns. A dark post frames the left edge; the veined white wall fills the background. Soft front-right light models her eyes and cheeks. From 00:00.000 to 00:02.200, she waits silently with tiny eyelid adjustments and steady shoulders. Keep the camera static.
[Shot 2] At 00:02.200, the camera cuts to <Subject 2>, seated on frame-right and facing screen-left in the lowered pose of <Picture 4>. He is already looking down-left at the cut. Soft front-left light reveals his face; dark timber remains at the right edge. At approximately 00:02.750-00:03.200, <Subject 2> (S1), in a low, measured adult male Mandarin voice, quietly answers: <d>[Chinese] 家。</d> His lips close naturally afterward; he holds his lowered chin until the cut.
[Shot 3] At 00:03.900, the camera cuts back to <Subject 1> in the same chest-up framing as [Shot 1]. Keep this ONE uninterrupted static shot through 00:13.500; the changes toward <Picture 2> and <Picture 3> are continuous acting, not edits or image morphs. From 00:03.900 to 00:04.250, she stares at him, slightly lifting her inner brows as if unsure she heard correctly. At approximately 00:04.250-00:06.200, <Subject 1> (S2), in a soft adult female Mandarin voice with a clear mid register and a trace of hurt, asks: <d>[Chinese] 家？你到底有几个家？</d> Briefly pause after the opening word, then finish the question in the same quiet voice. At 00:06.200-00:06.450 she takes a short breath without shifting her gaze.
At approximately 00:06.450-00:09.150, <Subject 1> (S2) suddenly raises her voice, tightens her brows and firms her shoulders toward the performance in <Picture 2>: <d>[Chinese] 你开全国连锁如家酒店的啊？！</d> Stress the chain-business and hotel-name phrases. Keep the voice recognizably hers, louder and sharper without audio clipping; her jaw and lips articulate naturally rather than freezing wide open. At 00:09.150-00:09.450 she takes a quick breath and begins leaning forward slightly from her seated torso.
At approximately 00:09.450-00:13.100, <Subject 1> (S2), now more indignant and leaning forward by about eight degrees toward the pose in <Picture 3>, demands: <d>[Chinese] 你是中国移动啊？！感情还全国覆盖吗？！</d> Give the first question until about 00:11.000, pause briefly to 00:11.200, then land the final challenge clearly. Her gaze stays on the man. Keep her anger sincere: no grin, laughter, tears, pointing, standing or cartoon distortion. By 00:13.100 she finishes, relaxes her mouth toward closure and holds the accusing eye line until the cut. Preserve face, hairpin, clothing and fixed background throughout.
[Shot 4] At 00:13.500, the camera cuts to <Subject 2> in the raised-eye reaction pose of <Picture 5>, using the same reverse-shot framing and lighting as [Shot 2]. He looks toward the woman, lips slightly parted as if about to answer, but produces no word or voiced sound. After a brief frozen beat, from about 00:13.900 to 00:14.650 he closes his lips and slowly lowers his eyes and chin toward the pose of <Picture 4>. Hold through 00:15.000. Keep his embarrassment small and deadpan, with no smile, big recoil or extra gesture.

overall_soundscape:
Low courtyard air and faint natural breathing continue across the cuts, with very subtle cloth movement. Keep the pauses audible and the louder speech clean. No additional voices, canned laughter, comic stings or exaggerated sound effects.

non_diegetic_music:
N/A



================================================================
prompt-024｜女主三句连续表演_10秒_I2VA
类型：视频　版本：全国覆盖（原资料版本） / H3提示词 / I2VA　历史记录：否
来源文件：女主三句连续表演_10秒_I2VA.txt
关联分镜：storyboard-05

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene, a ten-second uninterrupted 16:9 chest-up medium close-up beginning from the exact composition and mildly puzzled expression of <Picture 1>. The pictured adult woman occupies screen-left and faces the off-screen man to screen-right. Preserve her face shape, eyes, nose and lips, low black braided bun, plain gray hairpin, loose strands, white inner collar and silver-gray cross-collar robe with large ink-dyed areas. Broad continuous gray mineral veins remain attached to the luminous white stone wall, with dark timber at frame-left and soft front-right courtyard light. Skin retains faint natural warmth and fine texture. The camera holds a static shot at the same height and focal length throughout; there are no cuts, zooms, pans, reframing or focus pulls, and her hands remain outside the crop.
During 00:00.000-00:00.350, she silently holds the rightward eye line, lips slightly parted and brows showing quiet disbelief. From 00:00.350 to 00:02.300, the woman (S1), using a clear adult female Mandarin voice, says: <d>[Chinese] 家？你到底有几个家？</d> Deliver the first one-syllable question softly during approximately 00:00.350-00:00.650, pause briefly during approximately 00:00.650-00:00.850, then complete the remainder by 00:02.300. Her inner brows draw together slightly and the questioning tone becomes firmer, while her head and shoulders remain nearly still. Let the lips articulate every syllable naturally rather than holding a fixed open-mouth expression.
During 00:02.300-00:02.550, she takes a short breath without looking away. From 00:02.550 to 00:05.250, the same woman (S1) abruptly raises her volume and says: <d>[Chinese] 你开全国连锁如家酒店的啊？！</d> The eyebrows tighten, the eyes become more intent and the jaw opens only as required for clear syllables. Keep the delivery brisk but intelligible, with sharper emphasis on the chain-business comparison rather than equal shouting on every word. Her mouth returns toward a smaller opening at the end of the line. Preserve the same female vocal identity through the change in intensity.
During 00:05.250-00:05.550, she takes another short breath and begins a small continuous forward lean from the upper body, without a sudden head thrust, shoulder jerk or hand gesture. From 00:05.550 to 00:09.200, the woman (S1) delivers the final line with direct, escalating insistence: <d>[Chinese] 你是中国移动啊？！感情还全国覆盖吗？！</d> Speak the first question during approximately 00:05.550-00:07.100, hold a brief pause during 00:07.100-00:07.300, then complete the second question by 00:09.200. Keep her gaze on the man and the forward lean modest; allow only small brow and chin changes accompanying the words. The lips and jaw follow the Mandarin syllables with clear synchronized articulation, and they settle at the pauses instead of continuing speech-shaped motion.
From 00:09.200 to 00:10.000, she stops speaking, gently gathers her lips and holds the final questioning gaze and slight forward lean. Do not introduce another word, laugh, cry, turn away, look down or begin a new gesture. Maintain identity, hairpin position, garment markings, wall veins and lighting continuously across all three lines. The spoken comparisons do not introduce shops, phones, brand imagery or a change of setting. No subtitles, captions, titles, logos or other on-screen text appear anywhere. No other person enters and no off-screen reply is added.

overall_soundscape: Low courtyard air tone continues under the performance. Brief natural breathing accompanies the two pauses, with very subtle fabric movement during the small forward lean. No footsteps, dramatic effects or additional human voices occur.

non_diegetic_music: N/A



================================================================
prompt-025｜男主语塞收尾_4秒_I2VA
类型：视频　版本：全国覆盖（原资料版本） / H3提示词 / I2VA　历史记录：否
来源文件：男主语塞收尾_4秒_I2VA.txt
关联分镜：storyboard-08

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Live-action period scene, a four-second uninterrupted 16:9 chest-up male reverse shot beginning from the exact raised-gaze, speechless reaction in <Picture 1>. The pictured mature man occupies screen-right and faces the off-screen woman to screen-left. Preserve his face, straight dark brows, long eyes, natural nose and jaw, clean-shaven skin, complete black topknot and narrow dark band, dark-gray inner collar and silver-gray ink-dyed cross-collar robe. Keep the luminous white stone wall with broad continuous gray mineral veins, dark timber at frame-right and soft front-left courtyard light. Natural faint warmth remains in his skin, with readable eyes and continuous soft shadows beneath the brow. The camera holds a static shot at the same height and focal length throughout; there are no cuts, zooms, pans, reframing or focus pulls, and his hands remain outside the crop.
During 00:00.000-00:00.400, his eyes remain raised toward the off-screen woman and his lips are only slightly separated, matching the opening image. He appears momentarily unable to answer, but produces no spoken syllable, sigh, laugh or other vocalization. His head and shoulders hold nearly still; do not exaggerate the reaction into a wide-eyed grimace or a gaping mouth.
During 00:00.400-00:01.150, he slowly brings the lips together while lowering the eyes toward screen-left and down. The chin follows with a small continuous downward movement, leaving the upper body upright. The brows retain slight tension and the jaw settles without a smile, nod or abrupt head drop. Preserve the same eyes, face proportions and topknot through the motion.
From 00:01.150 to 00:04.000, maintain the closed-mouth, lowered-gaze state with only tiny natural facial adjustments and quiet breathing. The first 1.500 seconds form the intended reaction; the remaining time sustains the same pose as an editing tail without adding a new action. Do not look up again, speak, answer, shrug, turn away or introduce a hand gesture. Keep the robe's ink markings attached to the folds, the stone veins fixed to the wall and the light direction unchanged. No dialogue, voiceover, off-screen line or new person is added. No subtitles, captions, titles, logos or other on-screen text appear anywhere.

overall_soundscape: Low courtyard air tone and faint natural breathing continue throughout. Any fabric movement is very subtle. No vocal reactions, additional human voices, footsteps or dramatic effects occur.

non_diegetic_music: N/A



================================================================
prompt-026｜H3提示词_情绪递进_完整追问
类型：视频　版本：工作流修复_保留完整尾巴　历史记录：否
来源文件：H3提示词_情绪递进_完整追问.txt
关联分镜：storyboard-05、storyboard-06、storyboard-07

subject_definitions:
<Subject 1> is the same adult woman in <Picture 1>, <Picture 2> and <Picture 3>, with a low black braided bun, gray hairpin, loose side strands and a silver-gray ink-dyed cross-collar robe.
<Picture 1> guides the opening composition and puzzled expression in [Shot 1].
<Picture 2> guides her sudden indignant outburst in [Shot 1].
<Picture 3> guides her forward-leaning, confrontational expression in [Shot 1].
<Audio 1> is the voice-timbre reference for <Subject 1> (S1).

summary:
[reference generation + audio reference] A roughly 13-second, single-shot Mandarin comedy exchange. <Subject 1> challenges a man just off-screen, moving from quiet disbelief to an explosive accusation and a sarcastic final question. The three pictures guide her appearance and expressive progression; <Audio 1> maintains her recognizable voice.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain her face, hair and costume throughout the changing performance.
<Picture 1> ([Shot 1] opening composition): fully_preserved - retain her position, eyeline and pale stone setting.
<Picture 2> ([Shot 1] outburst): fully_preserved - use the expression as a performance reference within continuous movement.
<Picture 3> ([Shot 1] final challenge): fully_preserved - follow the forward posture and focused gaze.
<Audio 1>: reference - retain the speaker's timbre and Mandarin articulation while performing the new emotional progression and dialogue.

detailed_description:
Realistic live-action period comedy, pale stone with broad gray veining, dark timber and soft daylight. Natural skin texture and muted silver-gray fabric. No subtitles, captions or on-screen text.
[Shot 1] A steady chest-up shot places <Subject 1> left of center, looking at the man just off-screen right. Her face and shoulders remain clearly visible. The camera holds this composition through one continuous performance.

She starts speaking as the shot opens. Following <Picture 1>, her eyebrows lift unevenly and her head tilts, genuinely trying to understand his answer. <Subject 1> (S1), using the voice identity from <Audio 1>, asks softly but clearly, <d>[Chinese] 家？你到底有几个家？</d> The first word and the following question belong to one connected thought; a tiny questioning inflection leads directly onward. Disbelief builds across the sentence.

One short half-beat: her eyes widen as the implication lands, then her brows draw together. Following <Picture 2>, she lifts her chin and drives her shoulders forward. <Subject 1> (S1) suddenly projects her voice with sharp, indignant comic energy, <d>[Chinese] 你开全国连锁如家酒店的啊？！</d> Strong stress falls on the nationwide-chain phrase and the hotel name. Her eyes stay locked on him; an open palm turns toward him as if demanding an explanation. The outburst has clear consonants, supported breath and a forceful questioning finish.

That accusation immediately rolls into the final challenge, allowing only a quick breath. Following <Picture 3>, she leans closer, eyebrows lifted with incredulous sarcasm. <Subject 1> (S1) delivers both questions as one fluent burst, <d>[Chinese] 你是中国移动啊？！感情还全国覆盖吗？！</d> The mobile-operator question is a sharp accusation; the coverage question becomes a scathing rhetorical punchline. Emphasize the brand name and nationwide coverage, with changing pitch and facial expression rather than holding one uniform shout. Sustain the momentum through the last question and articulate its final syllable completely. Her hand gesture develops while she speaks, without creating extra silent beats.

Once the final syllable has fully sounded, she releases a short breath and holds a challenging, expectant look. Keep the spoken exchange flowing; the only deliberate dramatic pause is the half-beat before the outburst. Reserve a brief natural reaction after the completed speech.

overall_soundscape:
Clear synchronized female dialogue, natural breath and faint sleeve movement over quiet courtyard ambience. The off-screen man remains silent.

non_diegetic_music:
N/A



================================================================
prompt-027｜女主12秒_三图加声线参考_Ref2VA
类型：视频　版本：音色与情绪修正版　历史记录：否
来源文件：女主12秒_三图加声线参考_Ref2VA.txt
关联分镜：storyboard-05、storyboard-06、storyboard-07

subject_definitions:
<Subject 1> is the same adult woman across <Picture 1>, <Picture 2> and <Picture 3>: oval face, almond-shaped eyes, pale rose lips, low black braided bun, gray hairpin, loose side strands and a silver-gray ink-dyed cross-collar robe.
<Subject 2> is the shared classical courtyard: luminous white stone with broad gray mineral veins, dark timber at frame-left, gray paving and soft front-right daylight.
<Picture 1> guides the opening chest-up composition and quietly puzzled expression. The woman occupies frame-left and looks toward the off-screen man at screen-right.
<Picture 2> guides the later tightened brows and forceful speaking expression during the sudden outburst. It does not set her opening expression.
<Picture 3> guides the small seated forward lean during the final challenge. All three pictures show the same woman in successive states of ONE continuous shot.
<Audio 1> is the woman's voice reference extracted from the user's approved first clip. It defines the recognizable vocal identity of <Subject 1> (S1): her resonance, habitual vowel color, accent and articulation. Reference that voice for newly spoken dialogue; do not copy the source waveform, words, pauses or restrained emotional delivery.

summary:
[reference generation + audio reference] Create a 12-second, 16:9 live-action period-comedy performance by <Subject 1> in <Subject 2>, using the three pictures for visual continuity and <Audio 1> for the same woman's vocal identity. In one uninterrupted shot she starts genuinely puzzled and quiet, falls completely silent for a beat, suddenly bursts into a loud challenge, then leans forward and delivers a louder pointed follow-up. Make these three emotional levels clearly distinct while keeping the same voice. Use only the specified Chinese dialogue, with no subtitles.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain identity, hair, hairpin, layered collar and robe throughout the performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the fixed stone veins, dark timber, soft light and material colors.
<Picture 1> ([Shot 1] opening composition): partially_preserved - retain framing and eye line, allowing a softer, genuinely puzzled opening before anger appears.
<Picture 2> ([Shot 1] outburst): partially_preserved - apply its stronger brow tension and vocal articulation only after the silent beat, with natural moving lips rather than a frozen open mouth.
<Picture 3> ([Shot 1] final challenge): partially_preserved - apply the forward lean continuously near the final line, without changing the lens or face.
<Audio 1>: reference - preserve the same female vocal identity while changing loudness, pitch inflection, breath attack and emotional intensity as directed. Generate the new words; do not repeat source dialogue or carry its soft delivery across the whole target.

detailed_description:
One continuous live-action shot with clean white-gray-black physical materials, faint warm skin and natural fine facial texture. Cloth ink patterns remain attached to folds, and broad mineral veins stay fixed to the white wall. No subtitles, captions, titles, logos, visible writing, phones, modern props or hotel inserts.
[Shot 1] <Subject 1> sits on frame-left in the chest-up composition of <Picture 1>, facing the off-screen man to screen-right. Preserve the same face, low braided bun, gray hairpin, loose strands and silver-gray cross-collar robe. Dark timber stays at the left edge; the bright veined stone wall and gray paving remain behind her. Soft front-right light keeps her eyes and cheeks readable. The camera remains static throughout; her three emotional states develop continuously without cuts, image morphs or a zoom. Only one woman is visible.
From 00:00.000 to 00:00.400, she regards him in quiet disbelief. Her brow is relatively relaxed, inner brows slightly raised rather than deeply knitted, lips lightly parted and shoulders at rest. She has not started shouting or leaning forward.
At approximately 00:00.400-00:02.650, <Subject 1> (S1), in the same recognizable adult female Mandarin voice referenced from <Audio 1>, asks softly and sincerely: <d>[Chinese] 家？你到底有几个家？</d> Give the opening word a questioning rise, pause briefly, then finish the question at conversational volume. Keep her articulation intimate and clear, not already scolding. From 00:02.650 to 00:03.150, she stops speaking completely, gathers her lips, locks eyes with him and takes one quick preparatory breath. This half-second silence separates the soft question from the outburst.
At approximately 00:03.150-00:06.350, <Subject 1> (S1), retaining the same resonance and vowel color referenced from <Audio 1>, suddenly projects her voice much more strongly: <d>[Chinese] 你开全国连锁如家酒店的啊？！</d> The onset is abrupt and indignant, with a firm breath-supported attack. Her brows draw together toward the expression of <Picture 2>, eyes become intent and shoulders briefly tense. Stress the chain-business comparison and the hotel name, then land the final question sharply. This is a genuine angry raised voice, not a gently spoken line made louder afterward. Allow natural pitch movement and vocal edge while keeping the same woman recognizable; avoid a new cartoon voice or a shrill squeal. Her mouth opens and closes with the full syllables. At 00:06.350-00:06.700 she pauses to breathe, staying focused on him.
At approximately 00:06.700-00:10.950, <Subject 1> (S1), still using the voice identity from <Audio 1>, leans her seated torso forward slightly toward the state in <Picture 3> and demands: <d>[Chinese] 你是中国移动啊？！感情还全国覆盖吗？！</d> Begin the first question with incredulous emphasis; finish it around 00:08.450, pause briefly until about 00:08.700, then deliver the final challenge with unmistakable pressure on the nationwide-coverage phrase. Keep enough time for the final question particle to be fully audible. The anger remains sincere and the delivery forceful, without laughing at her own joke. Her head moves with the modest lean and natural speech emphasis; her gaze stays on the man. Do not hold her face and shoulders rigid while only moving the lips.
By approximately 00:10.950 she has completed the whole line. From there to 00:12.000, her lips close naturally and she holds the forward posture and accusing eye line in silence. Do not start another sentence or continue speech-shaped mouth movements. Preserve her face, hairpin, collar, fabric markings, light direction and fixed background. No pointing, standing, crying, off-screen response or new person.

overall_soundscape:
Quiet courtyard air, natural breaths at the pauses and a subtle cloth sound during the forward lean. Keep the soft opening intelligible and preserve the obvious dynamic change into the two louder lines, with no clipping, extra voices, canned laughter or comic sound effects.

non_diegetic_music:
N/A



================================================================
prompt-028｜07与08对比 / 00007_prompt 中保存的提示词
类型：视频　版本：07与08对比　历史记录：是
来源文件：00007_prompt.json
关联分镜：storyboard-05、storyboard-06、storyboard-07

subject_definitions:
<Subject 1> is the same adult woman in <Picture 1>, <Picture 2> and <Picture 3>, with the same face, low black braided bun, gray hairpin, loose side strands and silver-gray ink-dyed cross-collar robe.
<Picture 1> guides her opening composition and quiet, puzzled expression.
<Picture 2> guides her tightened brows and forceful expression during the sudden outburst.
<Picture 3> guides her small forward lean during the final challenge. These pictures depict ONE woman at successive stages of ONE continuous performance, not separate characters or separate shots.
<Audio 1> is the female voice from the user's first video and is the vocal-identity reference for <Subject 1> (S1). Preserve that woman's recognizable resonance, vowel color, accent and pronunciation. Generate new dialogue in her voice. Do not copy the source words, waveform, long pauses or restrained emotional delivery.

summary:
[reference generation + audio reference] Create a 12-second, 16:9 live-action period-comedy scene. In one uninterrupted shot, <Subject 1> changes from quiet disbelief to an abrupt angry outburst, then leans forward and delivers a pointed follow-up. Preserve the same female vocal identity from <Audio 1> throughout all three emotional stages. The contrast must be unmistakable: a genuinely soft opening, a brief silence, then a sudden forceful rise in voice and expression. Keep all specified Chinese dialogue complete and intelligible. No subtitles.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain her identity, hair, hairpin, robe and seated position throughout.
<Picture 1> ([Shot 1] opening): partially_preserved - retain composition and eye line, with a quiet, genuinely puzzled opening expression.
<Picture 2> ([Shot 1] outburst): partially_preserved - introduce the stronger expression only when the second line begins; animate natural speech articulation.
<Picture 3> ([Shot 1] final challenge): partially_preserved - develop the small forward lean continuously during the final line without changing camera position.
<Audio 1>: reference - preserve vocal identity while allowing natural changes in pitch, breath, loudness and vocal edge as anger increases. Do not transfer the source's calm emotional delivery to the entire target.

detailed_description:
Live-action cinematic realism in the same classical courtyard. Luminous white stone bears broad continuous gray mineral veins; dark timber frames the left edge. Her silver-gray robe has large ink-dyed areas attached to real folds. Soft front-right daylight models her face, with natural skin texture and faint warm skin color. Modern brand names remain spoken jokes only; no modern objects, signs, logos, subtitles or other visible text appear.

[Shot 1] <Subject 1> sits on frame-left in the chest-up composition of <Picture 1>, looking toward the off-screen man to screen-right. Keep the camera static and maintain the same background throughout. Her expression and posture change naturally in one continuous performance. Do not create cuts between the three picture references.

From 00:00.000 to 00:00.400, she looks at him as if checking whether she heard correctly. Her forehead is relatively relaxed, inner brows slightly raised, lips lightly parted and shoulders at rest. She is not already angry or shouting.

At approximately 00:00.400-00:02.650, <Subject 1> (S1), speaking in the same recognizable female Mandarin voice referenced from <Audio 1>, asks softly:
<d>[Chinese] 家？你到底有几个家？</d>
Give the opening word a small questioning rise, pause briefly, then finish the question at quiet conversational volume. Her expression carries disbelief and a trace of hurt. Keep her body upright and her gaze on him.

From 00:02.650 to 00:03.150, she stops speaking completely. Her lips gather, her eyes fix on him and she takes one quick preparatory breath. This half-second silence must clearly separate the soft question from the coming outburst.

At approximately 00:03.150-00:06.350, <Subject 1> (S1) suddenly raises her voice and delivers:
<d>[Chinese] 你开全国连锁如家酒店的啊？！</d>
Make the onset abrupt and forceful, with a firm breath-supported attack. Her brows draw together toward the expression in <Picture 2>, her eyes become intent and her shoulders briefly tense. Emphasize the chain-business comparison and the hotel name. She is sincerely angry, not performing a playful joke.
Preserve the recognizable voice identity from <Audio 1> while allowing a stronger natural pitch contour and vocal edge. Do not switch to a different woman, a cartoon voice or a thin artificial squeal. Her jaw and lips articulate the complete sentence naturally rather than remaining fixed in an open-mouth pose.

From 00:06.350 to 00:06.700, she takes a short breath without looking away. Keep this pause brief; do not insert a long empty gap.

At approximately 00:06.700-00:10.950, <Subject 1> (S1), still using the same vocal identity from <Audio 1>, leans her seated torso forward slightly toward the posture of <Picture 3> and demands:
<d>[Chinese] 你是中国移动啊？！感情还全国覆盖吗？！</d>
Deliver the first question with incredulous emphasis, finishing around 00:08.450. Pause briefly until approximately 00:08.700, then complete the final question by 00:10.950. Give the nationwide-coverage phrase strong, clear stress, and leave enough time to pronounce the final question particle fully. Do not rush or truncate the ending.
Her gaze remains locked on the man. Her head follows the small forward lean, and her face responds to the words with clear brow, cheek and jaw movement. Keep her anger sincere: no smile, laughter, crying or exaggerated comic grimace. The humor comes from the sudden emotional escalation and the unexpected dialogue.

From approximately 00:10.950 to 00:12.000, she finishes speaking, closes her lips naturally and holds the forward posture and questioning eye line in silence. Do not continue speech-shaped mouth movements or add another line. Keep this visible reaction tail.

Maintain the same face, hairpin, clothing, light direction and background. No pointing, standing, additional character, off-screen reply or camera movement. Preserve the contrast between the quiet first line and the two forceful later lines, with clean audio and no clipping.

overall_soundscape:
Quiet courtyard air, natural breaths during the pauses and subtle cloth movement during the forward lean. No additional human voices, canned laughter, dramatic music or comic sound effects.

non_diegetic_music:
N/A


================================================================
prompt-029｜07与08对比 / 00008_prompt 中保存的提示词
类型：视频　版本：07与08对比　历史记录：是
来源文件：00008_prompt.json
关联分镜：storyboard-05、storyboard-06、storyboard-07

subject_definitions:
<Subject 1> is the same adult woman in <Picture 1>, <Picture 2> and <Picture 3>, with the same face, low black braided bun, gray hairpin, loose side strands and silver-gray ink-dyed cross-collar robe.
<Picture 1> guides her opening composition and quiet, puzzled expression.
<Picture 2> guides her tightened brows and forceful expression during the sudden outburst.
<Picture 3> guides her small forward lean during the final challenge. These pictures depict ONE woman at successive stages of ONE continuous performance, not separate characters or separate shots.
<Audio 1> is the female voice from the user's first video and is the vocal-identity reference for <Subject 1> (S1). Preserve that woman's recognizable resonance, vowel color, accent and pronunciation. Generate new dialogue in her voice. Do not copy the source words, waveform, long pauses or restrained emotional delivery.

summary:
[reference generation + audio reference] Create a 13-second, 16:9 live-action period-comedy scene. In one uninterrupted shot, <Subject 1> changes from quiet disbelief to an abrupt angry outburst, then leans forward and delivers a pointed follow-up. Preserve the same female vocal identity from <Audio 1> throughout all three emotional stages. Make the contrast unmistakable: a genuinely soft opening, a brief silence, then a sudden forceful rise in voice and expression. Complete every word of the specified Chinese dialogue by 00:11.500, then leave a silent reaction through 00:13.000. No subtitles.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain her identity, hair, hairpin, robe and seated position throughout.
<Picture 1> ([Shot 1] opening): partially_preserved - retain composition and eye line, with a quiet, genuinely puzzled opening expression.
<Picture 2> ([Shot 1] outburst): partially_preserved - introduce the stronger expression only when the second line begins; animate natural speech articulation.
<Picture 3> ([Shot 1] final challenge): partially_preserved - develop the small forward lean continuously during the final line without changing camera position.
<Audio 1>: reference - preserve vocal identity while allowing natural changes in pitch, breath, loudness and vocal edge as anger increases. Do not transfer the source's calm emotional delivery to the entire target.

detailed_description:
Live-action cinematic realism in the same classical courtyard. Luminous white stone bears broad continuous gray mineral veins; dark timber frames the left edge. Her silver-gray robe has large ink-dyed areas attached to real folds. Soft front-right daylight models her face, with natural skin texture and faint warm skin color. Modern brand names remain spoken jokes only; no modern objects, signs, logos, subtitles or other visible text appear.

[Shot 1] <Subject 1> sits on frame-left in the chest-up composition of <Picture 1>, looking toward the off-screen man to screen-right. Keep the camera static and maintain the same background throughout. Her expression and posture change naturally in one continuous performance. Do not create cuts between the three picture references.

From 00:00.000 to 00:00.400, she looks at him as if checking whether she heard correctly. Her forehead is relatively relaxed, inner brows slightly raised, lips lightly parted and shoulders at rest. She is not already angry or shouting.

At approximately 00:00.400-00:02.650, <Subject 1> (S1), speaking in the same recognizable female Mandarin voice referenced from <Audio 1>, asks softly:
<d>[Chinese] 家？你到底有几个家？</d>
Give the opening word a small questioning rise, pause briefly, then finish the question at quiet conversational volume. Her expression carries disbelief and a trace of hurt. Keep her body upright and her gaze on him. Her lips articulate naturally and settle during the short pause.

From 00:02.650 to 00:03.150, she stops speaking completely. Her lips gather, her eyes fix on him and she takes one quick preparatory breath. This half-second silence clearly separates the soft question from the coming outburst.

At approximately 00:03.150-00:06.350, <Subject 1> (S1) suddenly raises her voice and delivers:
<d>[Chinese] 你开全国连锁如家酒店的啊？！</d>
Make the onset abrupt and forceful, with a firm breath-supported attack. Her brows draw together toward the expression in <Picture 2>, her eyes become intent and her shoulders briefly tense. Emphasize the chain-business comparison and the hotel name. She is sincerely angry, not performing a playful joke.
Preserve the recognizable voice identity from <Audio 1> while allowing a stronger natural pitch contour and vocal edge. Do not switch to a different woman, a cartoon voice or a thin artificial squeal. Her jaw and lips articulate the complete sentence naturally rather than remaining fixed in an open-mouth pose. Finish the final syllable clearly.

From 00:06.350 to 00:06.700, she takes a short breath without looking away. Keep this pause brief; do not insert a long empty gap.

At approximately 00:06.700-00:11.500, <Subject 1> (S1), still using the same vocal identity from <Audio 1>, leans her seated torso forward slightly toward the posture of <Picture 3> and demands:
<d>[Chinese] 你是中国移动啊？！感情还全国覆盖吗？！</d>
Deliver the first question with incredulous emphasis, finishing around 00:08.450. Pause briefly until approximately 00:08.700, then deliver the second question with firm, clear stress on the nationwide-coverage phrase.
Keep a steady, intelligible pace through the final question. Do not accelerate the last words to fit the ending. Reserve enough articulation time for the final “吗”, including its complete vowel and questioning intonation. It must be fully audible as the natural ending of the sentence, not swallowed, whispered away, clipped or separated into an extra repeated word. Complete the whole sentence by 00:11.500.
Her gaze remains locked on the man. Her head follows the small forward lean, and her face responds to the words with clear brow, cheek and jaw movement. Keep her anger sincere: no smile, laughter, crying or exaggerated comic grimace. The humor comes from the sudden emotional escalation and the unexpected dialogue.

From 00:11.500 to 00:13.000, she has completely finished speaking. Her mouth closes naturally after the final vowel, and she holds the forward posture and questioning eye line in silence. Preserve this full silent reaction tail. Do not add another word, repeat the last syllable or continue speech-shaped mouth movements. Do not stretch the dialogue across the final silent interval.

Maintain the same face, hairpin, clothing, light direction and background. No pointing, standing, additional character, off-screen reply or camera movement. Preserve the contrast between the quiet first line and the two forceful later lines, with clean audio and no clipping.

overall_soundscape:
Quiet courtyard air, natural breaths during the pauses and subtle cloth movement during the forward lean. The final silent reaction contains only faint natural breathing and courtyard ambience. No additional human voices, canned laughter, dramatic music or comic sound effects.

non_diegetic_music:
N/A


================================================================
prompt-030｜09复核 / prompt 中保存的提示词
类型：视频　版本：09复核　历史记录：是
来源文件：prompt.json
关联分镜：storyboard-05、storyboard-06、storyboard-07

subject_definitions:
<Subject 1> is the same adult woman in <Picture 1>, <Picture 2> and <Picture 3>, preserving her face, low black braided bun, gray hairpin, loose side strands and silver-gray ink-dyed cross-collar robe.
<Picture 1> guides her opening composition and quietly puzzled expression.
<Picture 2> guides the stronger expression during her sudden angry outburst.
<Picture 3> guides her forward lean and direct questioning near the end. All three pictures depict ONE woman within ONE continuous shot.
<Audio 1> is the vocal-identity reference for <Subject 1> (S1). Preserve the same woman's recognizable resonance, vowel color, accent and pronunciation. Her speaking pace, pauses and emotional intensity follow the target performance below, not the calm delivery of the reference recording.

summary:
[reference generation + audio reference] Create a 13-second, 16:9 live-action period-comedy performance. <Subject 1> begins with quiet disbelief, pauses briefly, then suddenly bursts into an angry challenge and follows through with two connected questions. Use <Audio 1> for her vocal identity while allowing a forceful, expressive performance. Complete the dialogue within approximately the first 10.5 seconds, leaving the remaining time for her silent reaction. No subtitles.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain her identity, hair, hairpin and clothing.
<Picture 1> ([Shot 1] opening): partially_preserved - retain framing and eye line while animating quiet disbelief.
<Picture 2> ([Shot 1] outburst): partially_preserved - use its tightened brows and forceful expression after the opening question.
<Picture 3> ([Shot 1] final challenge): partially_preserved - develop its forward lean naturally during the final questions.
<Audio 1>: reference - retain the same female vocal identity, but generate the specified new dialogue with stronger projection, sharper emphasis and a brisker questioning rhythm. Do not copy the reference words, waveform, pauses or emotional restraint.

detailed_description:
Live-action cinematic realism in the established classical courtyard. Broad gray mineral veins remain attached to luminous white stone. Dark timber frames the left edge, and large ink-dyed areas follow the folds of her silver-gray robe. Soft front-right daylight preserves natural facial texture and faint warm skin color. Modern brands exist only in the spoken dialogue. No subtitles, captions, logos or modern objects appear.

[Shot 1] <Subject 1> sits on frame-left in the chest-up composition of <Picture 1>, looking toward the off-screen man to screen-right. The camera remains static throughout. Her changing expression, speech and posture form one continuous performance, without cuts between the reference pictures.

Begin speaking promptly, with only a brief opening look. During approximately the first 2.2 seconds, <Subject 1> (S1), using the recognizable female Mandarin voice referenced from <Audio 1>, asks quietly:
<d>[Chinese] 家？你到底有几个家？</d>
The opening word carries genuine disbelief. Give it a brief questioning beat, then continue the sentence naturally. Her eyes search his face and her inner brows rise slightly. She is still trying to understand his answer, not already shouting.

After the question, allow ONE short pause of about 0.3 seconds. She gathers her lips, fixes her gaze on him and takes a quick breath.

Beginning around 00:02.500, <Subject 1> (S1) suddenly projects her voice and angrily challenges him:
<d>[Chinese] 你开全国连锁如家酒店的啊？！</d>
Finish this line around 00:05.400. Make the vocal attack abrupt and clearly stronger than the opening. Her brows tighten toward the expression in <Picture 2>, her jaw articulates firmly and her head gives a small natural emphasis with the words. Stress the chain-business comparison and hotel name. She is sincerely furious; she does not smile at the joke. Keep her voice recognizably the same woman while allowing natural pitch changes and a stronger vocal edge.

Take only a quick breath, then continue immediately. Beginning around 00:05.600, <Subject 1> (S1) leans forward toward the posture of <Picture 3> and demands:
<d>[Chinese] 你是中国移动啊？！</d>
Finish this question around 00:07.400. Her eyes stay fixed on him and her expression becomes more pointed. The movement comes naturally from the questioning, with her head and shoulders participating rather than remaining rigid.

After a very brief beat of about 0.2 seconds, <Subject 1> (S1) continues the same angry thought:
<d>[Chinese] 感情还全国覆盖吗？！</d>
Begin this final question around 00:07.600 and give it enough space to finish clearly by approximately 00:10.500. It is a direct follow-up, not a new speech after a long pause. Maintain a clear, even pace through the entire question. Emphasize the nationwide-coverage phrase and fully pronounce the final question particle with its complete vowel and questioning intonation. Do not accelerate, trail away or swallow the last syllable.

These timing cues organize the delivery; do not insert extra silence between the last three lines. The important rhythm is: quiet question, one short pause, sudden outburst, quick breath, two closely connected challenges.

After the final question has been fully spoken, her lips close naturally. For the remaining time through 00:13.000, she holds her forward posture and looks at him, waiting for an answer. This is a silent reaction after completed speech. No additional words, repeated syllables or continued speaking-shaped mouth movement.

Keep the performance expressive and physically natural. Her anger develops through her eyes, brows, jaw, breath and forward lean. Preserve her face, hairpin, robe, lighting and background throughout. No additional character or off-screen reply.

overall_soundscape:
Quiet courtyard air, natural short breaths between phrases and subtle cloth movement during the forward lean. Preserve the audible contrast between the quiet opening and the forceful later dialogue. No additional voices, canned laughter or comic sound effects.

non_diegetic_music:
N/A


================================================================
prompt-031｜爆发开骂
类型：生图　版本：女主回怼_报战绩　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-13

生成一张独立完整的16:9横向真人电影分镜参考图。只有一位成年女子，固定胸上近景，眼平机位，女主位于画面左侧，身体和脸朝画外右侧的男子，右侧保留视线空间，不看镜头；发髻和发饰完整可见，双手留在画外。输入图1负责身份、服装、光线、场景和基本构图，输入图2辅助同脸与身体前倾姿态。延续输入图中同一个成年女主：自然清秀椭圆脸，横向舒展的杏眼，细直鼻梁，自然鼻尖，唇峰清楚的淡玫瑰色嘴唇，真实细微肤理和淡暖肤色；黑发梳成低位编织发髻，素灰簪，两颊少量松散发丝。银灰交领长衣在胸肩处层叠垂坠，大块浓淡墨染依附真实衣褶。她身后是同一古典庭院的明亮白石墙，带连续宽幅灰色石脉，左后方深色木柱，灰石地面。侧前方宽幅自然柔光塑造脸部连续体积，环境白灰黑，肤色微暖。此刻她正在回怼刚刚炫耀感情经历的男子，要有喜剧里的真生气和带刺讥讽，不能演成害怕、委屈落泪或温柔含情。这一帧捕捉她突然开骂的有力瞬间：身体暂时端正，眉心骤然收紧，眉头内侧压低，眼神锐利锁住画外右侧，眼睛保持清楚可读，嘴自然张开发声，下颌有支撑而不夸张脱位，脸部肌肉和颈侧带出瞬间的力度。她对男人的得意明显恼火。嘴张开程度比参考图更明确，不能像惊吓尖叫，不露异常牙齿，不流泪。保持同一个女主的骨相、年龄、发髻、衣纹、白石纹和固定机位。画面没有字幕、对白文字、编号、边框、水印、分屏、拼图；不出现男主或其他人物，不添加手势、武器或物理攻击，不做卡通夸张大眼、塑料磨皮、胶片粗粒或水墨插画。


================================================================
prompt-032｜前倾逼问
类型：生图　版本：女主回怼_报战绩　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-14

生成一张独立完整的16:9横向真人电影分镜参考图。只有一位成年女子，固定胸上近景，眼平机位，女主位于画面左侧，身体和脸朝画外右侧的男子，右侧保留视线空间，不看镜头；发髻和发饰完整可见，双手留在画外。输入图1负责身份、服装、光线、场景和基本构图，输入图2辅助同脸与身体前倾姿态。延续输入图中同一个成年女主：自然清秀椭圆脸，横向舒展的杏眼，细直鼻梁，自然鼻尖，唇峰清楚的淡玫瑰色嘴唇，真实细微肤理和淡暖肤色；黑发梳成低位编织发髻，素灰簪，两颊少量松散发丝。银灰交领长衣在胸肩处层叠垂坠，大块浓淡墨染依附真实衣褶。她身后是同一古典庭院的明亮白石墙，带连续宽幅灰色石脉，左后方深色木柱，灰石地面。侧前方宽幅自然柔光塑造脸部连续体积，环境白灰黑，肤色微暖。此刻她正在回怼刚刚炫耀感情经历的男子，要有喜剧里的真生气和带刺讥讽，不能演成害怕、委屈落泪或温柔含情。这一帧捕捉紧接开骂之后的逼问：上身从胸肩处明显但自然地向画外右侧前倾，脖子与身体一起向前而不凭空伸长；下巴略收、眉毛压低，目光直接而有压迫感。嘴唇处于自然半张的连续问话口型，表情是在质问对方有什么可得意的。和第一张相比，身体前倾、嘴形收窄、视线更集中，保持愤怒和清醒。保持同一个女主的骨相、年龄、发髻、衣纹、白石纹和固定机位。画面没有字幕、对白文字、编号、边框、水印、分屏、拼图；不出现男主或其他人物，不添加手势、武器或物理攻击，不做卡通夸张大眼、塑料磨皮、胶片粗粒或水墨插画。


================================================================
prompt-033｜冷讽挑眉
类型：生图　版本：女主回怼_报战绩　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-15

生成一张独立完整的16:9横向真人电影分镜参考图。只有一位成年女子，固定胸上近景，眼平机位，女主位于画面左侧，身体和脸朝画外右侧的男子，右侧保留视线空间，不看镜头；发髻和发饰完整可见，双手留在画外。输入图1负责身份、服装、光线、场景和基本构图，输入图2辅助同脸与身体前倾姿态。延续输入图中同一个成年女主：自然清秀椭圆脸，横向舒展的杏眼，细直鼻梁，自然鼻尖，唇峰清楚的淡玫瑰色嘴唇，真实细微肤理和淡暖肤色；黑发梳成低位编织发髻，素灰簪，两颊少量松散发丝。银灰交领长衣在胸肩处层叠垂坠，大块浓淡墨染依附真实衣褶。她身后是同一古典庭院的明亮白石墙，带连续宽幅灰色石脉，左后方深色木柱，灰石地面。侧前方宽幅自然柔光塑造脸部连续体积，环境白灰黑，肤色微暖。此刻她正在回怼刚刚炫耀感情经历的男子，要有喜剧里的真生气和带刺讥讽，不能演成害怕、委屈落泪或温柔含情。这一帧捕捉怒气转成带刺讽刺的瞬间：延续前倾，头稍稍偏转但仍看画外右侧，一侧眉毛抬起、另一侧仍带怒意，下巴微抬，嘴角出现极轻的一边冷笑，同时双唇自然微张正在说挖苦的话。她被对方气得想笑，眼神却清楚地不屑，冷笑没有亲切或调情意味。与前一张直冲的责问形成可读的表情反差，面部比例自然。保持同一个女主的骨相、年龄、发髻、衣纹、白石纹和固定机位。画面没有字幕、对白文字、编号、边框、水印、分屏、拼图；不出现男主或其他人物，不添加手势、武器或物理攻击，不做卡通夸张大眼、塑料磨皮、胶片粗粒或水墨插画。


================================================================
prompt-034｜收口冷眼
类型：生图　版本：女主回怼_报战绩　历史记录：否
来源文件：分镜与生图提示词.json
关联分镜：storyboard-16

生成一张独立完整的16:9横向真人电影分镜参考图。只有一位成年女子，固定胸上近景，眼平机位，女主位于画面左侧，身体和脸朝画外右侧的男子，右侧保留视线空间，不看镜头；发髻和发饰完整可见，双手留在画外。输入图1负责身份、服装、光线、场景和基本构图，输入图2辅助同脸与身体前倾姿态。延续输入图中同一个成年女主：自然清秀椭圆脸，横向舒展的杏眼，细直鼻梁，自然鼻尖，唇峰清楚的淡玫瑰色嘴唇，真实细微肤理和淡暖肤色；黑发梳成低位编织发髻，素灰簪，两颊少量松散发丝。银灰交领长衣在胸肩处层叠垂坠，大块浓淡墨染依附真实衣褶。她身后是同一古典庭院的明亮白石墙，带连续宽幅灰色石脉，左后方深色木柱，灰石地面。侧前方宽幅自然柔光塑造脸部连续体积，环境白灰黑，肤色微暖。此刻她正在回怼刚刚炫耀感情经历的男子，要有喜剧里的真生气和带刺讥讽，不能演成害怕、委屈落泪或温柔含情。这是所有台词已经完整说完后的结尾一帧：双唇自然合拢成平直清楚的唇线，冷笑消失，下巴稍稍压下，仍保持些许前倾；眉心比开骂时略松，眼神却冷而锐利地锁住画外右侧，像在等对方再敢狡辩。身体稳定、有余怒，没有新的说话口型，没有笑容，没有哭意。要能和第三张的冷讽说话口型清楚区分。保持同一个女主的骨相、年龄、发髻、衣纹、白石纹和固定机位。画面没有字幕、对白文字、编号、边框、水印、分屏、拼图；不出现男主或其他人物，不添加手势、武器或物理攻击，不做卡通夸张大眼、塑料磨皮、胶片粗粒或水墨插画。


================================================================
prompt-035｜H3提示词_男主两省一市_8秒_Ref2VA
类型：视频　版本：两省一市_男主结尾　历史记录：否
来源文件：H3提示词_男主两省一市_8秒_Ref2VA.txt
关联分镜：storyboard-09、storyboard-10、storyboard-11、storyboard-12

subject_definitions:
<Subject 1> is the same mature man in <Picture 1>, <Picture 2>, <Picture 3> and <Picture 4>, with a black topknot, clean-shaven face and silver-gray ink-dyed cross-collar robe.
<Picture 1> guides the opening composition and feigned innocence in [Shot 1].
<Picture 2> guides his emerging smug smile in [Shot 1].
<Picture 3> guides his self-satisfied speaking expression in [Shot 1].
<Picture 4> guides his final closed-mouth smirk in [Shot 1].

summary:
[reference generation] An 8-second, single-shot Mandarin comedy ending. <Subject 1> answers a woman off-screen, pretending to correct an exaggeration, then inadvertently boasting about his romantic reach. His performance develops continuously through the four picture references.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain his face, mature appearance, topknot and costume.
<Picture 1> ([Shot 1] opening): fully_preserved - retain the framing, eyeline and pale stone setting.
<Picture 2> ([Shot 1] second phrase): fully_preserved - follow the raised gaze and emerging smile.
<Picture 3> ([Shot 1] place-name list): fully_preserved - follow the relaxed, proud speaking expression.
<Picture 4> ([Shot 1] ending): fully_preserved - follow the narrowed eyes and asymmetric closed-mouth smirk.

detailed_description:
Realistic live-action period comedy. Broad gray veins cross the luminous white stone background; dark timber frames the right edge. Soft daylight shapes natural, faintly warm skin and the robe's flowing ink patterns. No subtitles, captions, logos or on-screen text.

[Shot 1] A locked chest-up shot holds <Subject 1> on the right, facing the woman off-screen left, matching <Picture 1>. His topknot remains fully visible and his hands stay outside the frame. He begins speaking immediately, with a natural, resonant adult male voice in a low-to-middle register and clear conversational Mandarin.

<Subject 1> (S1) draws his inner eyebrows together, lowers his chin slightly and answers with mock innocence, <d>[Chinese] 哪有全国。</d> He sounds mildly aggrieved, as though she has unfairly exaggerated. His gaze remains on her.

His expression changes while he continues speaking. Following <Picture 2>, his brow relaxes, his chin rises a little and one corner of his mouth begins to lift. <Subject 1> (S1) replies with casual, self-satisfied certainty, <d>[Chinese] 也就两省一市。</d> Underplay the opening words as if this were a modest achievement, then give the geographical count a precise, matter-of-fact emphasis.

He flows directly into the list, following <Picture 3>. <Subject 1> (S1) says with growing, barely concealed pride, <d>[Chinese] 吉林、天津、浙江。</d> The three names form one connected phrase with brief comma separations. His eyes brighten and the smile spreads as he speaks. His face and breath carry the comedy; the voice remains the same male voice throughout. Keep the exchange moving at a natural conversational pace, allowing the spoken lines to finish in roughly the first six seconds.

Only after the last place-name is fully pronounced does <Subject 1> (S1) release one short, low, breathy chuckle. Following <Picture 4>, he closes his lips, lets one mouth corner rise higher and narrows his eyes in unmistakable smug amusement. He looks pleased with his own explanation. Retain a brief complete reaction after the chuckle, holding his final smirk through the end of the shot.

overall_soundscape:
Clear synchronized male dialogue, one brief male chuckle, natural breathing and faint courtyard ambience. The woman remains silent off-screen.

non_diegetic_music:
N/A



================================================================
prompt-036｜H3提示词_女主回怼_情绪递进_9秒_Ref2VA
类型：视频　版本：女主回怼_报战绩　历史记录：否
来源文件：H3提示词_女主回怼_情绪递进_9秒_Ref2VA.txt
关联分镜：storyboard-13、storyboard-14、storyboard-15、storyboard-16

subject_definitions:
<Subject 1> is the same adult woman in <Picture 1>, <Picture 2>, <Picture 3> and <Picture 4>, with a low braided black bun, gray hairpin and silver-gray ink-dyed cross-collar robe.
<Picture 1> guides her opening outburst in [Shot 1].
<Picture 2> guides her forward-leaning confrontation in [Shot 1].
<Picture 3> guides her sarcastic speaking expression in [Shot 1].
<Picture 4> guides her silent, closed-mouth stare at the end of [Shot 1].
<Audio 1> is the voice-timbre reference for <Subject 1> (S1).

summary:
[reference generation + audio reference] A 9-second, single-shot Mandarin comedy reaction. <Subject 1> responds to an off-screen man's smug romantic boasting with an angry insult, an incredulous challenge and biting sarcasm. The four pictures guide her continuous performance; <Audio 1> preserves her recognizable voice.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain her identity, hairstyle and costume.
<Picture 1> ([Shot 1] opening): fully_preserved - retain composition, eyeline, setting and the angry speaking expression.
<Picture 2> ([Shot 1] second phrase): fully_preserved - follow the forward posture and confrontational gaze.
<Picture 3> ([Shot 1] final phrase): fully_preserved - follow the slight head tilt and sarcastic expression.
<Picture 4> ([Shot 1] ending): fully_preserved - follow the closed lips and cold stare after all speech.
<Audio 1>: reference - preserve vocal identity, timbre and natural accent while performing the new dialogue and emotional progression.

detailed_description:
Realistic live-action period comedy. Luminous white stone with broad gray veins, dark timber, silver-gray ink-dyed fabric and soft directional daylight surround naturally warm skin. No subtitles or on-screen text.

[Shot 1] A locked chest-up shot holds <Subject 1> left of center, facing the man off-screen right. Her bun remains visible and her hands stay below frame. The composition follows <Picture 1>. She is already looking at him and starts speaking immediately. Throughout the shot, <Audio 1> supplies her vocal identity; breath pressure, pitch, emphasis and facial movement express the changing anger.

<Subject 1> (S1) snaps, <d>[Chinese] 你个瘪犊子，</d> Her brows contract sharply, her eyes fix on him and her jaw opens decisively. The insult bursts out on a firm, supported breath, with crisp consonants and a strong stress at its end. She sounds personally offended by his shameless pride. Her shoulders tense with the words; her voice is full and forceful.

Following <Picture 2>, <Subject 1> (S1) immediately leans toward him while continuing, <d>[Chinese] 搁这儿报战绩呢？</d> She lowers her pitch slightly, compressing the anger into a pointed accusation. Her brows press down and her chin moves forward. She gives the phrase about boasting of achievements a hard, incredulous emphasis, then lifts the final questioning inflection. The change from a sudden outburst to a close, challenging delivery is clearly audible and visible.

A quick breath carries directly into the final phrase. Following <Picture 3>, <Subject 1> (S1) tilts her head slightly and lets one mouth corner curl with scorn, saying, <d>[Chinese] 还挺会开疆拓土啊？！</d> The opening carries exaggerated false admiration, as though she cannot believe he expects praise. Her eyebrow lifts; the smile remains sharp with anger. As she reaches the accusation about expanding territory, her articulation becomes firmer and her voice projects toward him again. The final questioning vowel is fully voiced and clearly audible. Her facial movement and emphasis develop during the sentence, without a separate laugh or silent pose.

Keep the three phrases connected at a natural, energetic speaking pace, completing all dialogue within roughly the first 7.5 seconds. Every word remains intelligible, including the last vowel. Only after it finishes does she close her lips and settle into <Picture 4>: chin slightly lowered, eyes cold and unwavering. Hold this challenging reaction for the remaining time, with a small natural exhale. The man stays silent off-screen.

overall_soundscape:
Quiet courtyard ambience, close natural breathing and faint fabric movement. The dialogue remains clear and synchronized with her mouth.

non_diegetic_music:
N/A