《玩偶之家》

用途：identity-preserve，电影对白二创的独立分镜照片。生成一张16:9横屏写实电影静帧，单张完整画面，不是拼图。参考图是本项目已确定的双人定妆照，作为唯一的人物身份、服装、空间和光线连续性锚点。
锁定人物：恰好两名成年人。画面左侧是参考中成熟东亚男性，保留其具体脸型、鼻梁、略显疲惫的眼睛、蓬松微乱短黑发与淡胡茬，象牙白卷袖亚麻衬衫、米色长裤、深色腰带。画面右侧是参考中的同一成年东亚女性，保留其具体五官、低盘髻和鬓边碎发，奶油白闭合立领褶皱长袖衬衫、深酒红高腰长裙。不得年轻化男主，不得替换女主的脸，不得改变衣领、纽扣和服装，不添加首饰。
空间锁定：两人在同一间植物书房的同一小块地面近距离面对面，相距约60至70厘米，始终男左女右，眼神相向。不是分别站在走廊两端，也不是两张单人照片拼接。两人之间没有门柱、桌子或隔断。相机留在原定妆图所示的对话轴线同一侧，可改变景别而不能反转人物左右。
风格锁定：「混合宅邸·精致暗彩」。墨绿墙、深胡桃木护墙和背景实体雕花门套形成一至两层纵深，局部细花纹与素面交替；右侧植物纹彩绘玻璃窗投来柔和日光，深木压住奶油白衣亮部，少量梅紫酒红织物。与参考保持同一时刻同一光向，暗部可辨，肤色真实，白衣不过曝。人物面部落在安静的背景上，门框均在身后，不穿过头或身体。环境现有植物标本册、玻璃植物箱、黄铜灯只可在取景包含时出现，不增添新道具。真实材质与细小肤纹，高质量电影剧照，不额外添加胶片颗粒、扫描痕迹或噪点。
画面里只能出现此男女两人。无字幕、无对话气泡、无文字、无编号、无水印、无手机界面。没有拥抱、亲吻、触碰对方、指戳脸或肢体冲突。表演应真实鲜活，不是广告摆拍；除非本镜头另有描述，正在说话的人嘴略张、听话者闭嘴，不让两人同时夸张张嘴。双手自然，避免遮脸、交叉或多指。
本镜头：第1—2轮对白的开场关键帧。腰部以上的双人中景，两人完整头部、肩膀和手部均在画面中，人物比原定妆照更紧凑地占画面。男主左侧三分之二侧脸清楚，身体挺直，正以理所当然的态度说话：眉峰轻扬、口型自然略开，一只手在腰间掌心朝上向前摊开，另一只手自然垂着。女主右侧三分之二侧脸清楚，听到“母亲、妻子”后有短暂受伤的表情：眉心轻拢，下唇轻抿，眼睛略有水光但没有眼泪，下巴仍平稳；双手在腰前轻握，指尖略紧，正在抬眼迎上他的目光。用两个人不同的微表情表现权力不对等和将要反驳的蓄势，不能两个人同一张冷脸。背景保留连续木门框以及右侧一小段彩绘玻璃，明暗方向照参考。

subject_definitions:
<Subject 1> is the adult man on the left in <Picture 1>, with tousled dark hair, light stubble, an ivory shirt with rolled sleeves, beige trousers and a dark belt.
<Subject 2> is the adult woman on the right in <Picture 1>, with a low dark bun, an ivory high-neck long-sleeved blouse and a muted burgundy long skirt.
<Subject 3> is the mansion study in <Picture 1>, with deep green walls, carved dark walnut, layered doorways, stained-glass window light and a wooden desk.
<Picture 1> is the composition and performance-blocking reference for [Shot 1], placing both characters close together in one frame.
<Audio 1> is the male voice-identity reference for <Subject 1> (S1).
<Audio 2> is the female voice-identity reference for <Subject 2> (S2).

summary:
[reference generation + audio reference] The target video is a 13-second confrontation in which the man shouts the wife's prescribed role and the woman fiercely asserts her equal personhood, using <Picture 1> for <Subject 1>, <Subject 2>, <Subject 3> and their close two-shot; <Audio 1> and <Audio 2> provide the separate male and female voice identities.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the man's facial identity, hair, apparent age and clothing throughout.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the woman's facial identity, hair, apparent age and clothing throughout.
<Subject 3> (appears in [Shot 1]): fully_preserved - retain the mansion study, materials, palette and window-light direction.
<Picture 1> ([Shot 1] composition reference): fully_preserved - preserve the close two-person arrangement and viewpoint while allowing the specified acting.
<Audio 1>: reference - retain the male speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.
<Audio 2>: reference - retain the female speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.

detailed_description:
The target video is a 16:9 realistic dramatic film scene with deep green walls, dark carved wood, subdued burgundy fabric and ivory clothing. Soft directional window light shapes natural faces, while the intense argument remains the focus; there are no subtitles, captions, logos or additional people.
Both voices remain recognizable as their respective audio references. At the specified peaks, anger produces a loud projected shout with a raw, hoarse, grainy edge and a brief strained vocal break; pitch and roughness may rise naturally without changing the speaker's identity. This is an audible shouted performance with clear syllables, not a whisper, a breathy hiss, a calmly recited sentence or merely a louder recording. Preserve clean microphone sound without digital distortion.
[Shot 1] A steady eye-level medium two-shot follows the composition of <Picture 1> inside <Subject 3>. <Subject 1> stands on the left and <Subject 2> on the right, about sixty centimeters apart, both faces visible in three-quarter view. Keep the frame close enough to read their mouths and eyes; hold the camera still during the vocal peaks.
During approximately the first four seconds, <Subject 1> (S1) uses the male voice identity from <Audio 1>. He is furious that she is challenging the authority he assumes over her. With a hard stare and a rigid jaw, he drives the opening clause at her in an already raised, biting voice, stressing the priority he is imposing: <d>[Chinese] 你首先是一个母亲，</d>
He catches one short breath at the comma, about a fifth of a second, and the anger breaks through his control. <Subject 1> (S1) SHOUTS THE ENTIRE NEXT CLAUSE with forceful chest projection and a raw, hoarse, cracking edge: <d>[Chinese] 是一个妻子。</d> The burst begins on the very first syllable of this clause, builds through its connecting words, and peaks on the final two-syllable word meaning wife. Put the strongest attack on its first syllable and keep the last syllable fully voiced and intelligible before the sound releases. His volume and pitch jump audibly above his first clause; he does not trail off into a conversational ending. His jaw drops, his brows press inward and his torso pitches slightly toward her as one firm downward hand beat lands with the vocal peak. Keep his recognizable male timbre through the temporary rasp. Both clauses form one continuous accusation: speak them once, in order, without an added word, a long pause or a wordless scream. After the final syllable, he closes his mouth and holds his glare while she answers.
<Subject 2> receives the shout with a brief blink, a tightened jaw and a sharply lifted chin. After only a short turn-taking gap, <Subject 2> (S2), using the female voice identity from <Audio 2>, releases one compact forearm gesture and answers directly into his face. She says exactly once: <d>[Chinese] 我首先是一个人，一个和你一样的人。</d> The first occurrence of ren, meaning person, erupts in a loud, hoarse, full-voice shout. At the comma she takes only a quick catch of breath; she does not return to calm speech. The second clause drives forward, stressing her equality with him, then the second and final ren is shouted at least as forcefully as the first, with a rough breaking edge and complete, intelligible voicing. Her brows draw together, her eyes stay on him, and her open mouth and small forward torso movement coincide with both stressed words. This is anger at being denied dignity; she does not smile or laugh. He stops his hand gesture as she answers and keeps his mouth closed.
Aim to finish the spoken exchange around ten to eleven seconds, leaving the remaining time for her natural exhale and their tense eye contact. Keep any additional breaths brief and allow the final ren to finish before her lips close; do not stretch internal pauses or slow the second clause into a quiet declaration. Hold the same composition to the end without adding dialogue or fading the audio early.

overall_soundscape:
Low indoor room tone, light clothing movement and natural breaths remain underneath the dialogue. The projected voices dominate the sound field with a short, natural indoor reflection. No competing voices, laughter or sound effects cover the final syllables.

non_diegetic_music:
N/A

用途：identity-preserve，电影对白二创的独立分镜照片。生成一张16:9横屏写实电影静帧，单张完整画面，不是拼图。参考图是本项目已确定的双人定妆照，作为唯一的人物身份、服装、空间和光线连续性锚点。
锁定人物：恰好两名成年人。画面左侧是参考中成熟东亚男性，保留其具体脸型、鼻梁、略显疲惫的眼睛、蓬松微乱短黑发与淡胡茬，象牙白卷袖亚麻衬衫、米色长裤、深色腰带。画面右侧是参考中的同一成年东亚女性，保留其具体五官、低盘髻和鬓边碎发，奶油白闭合立领褶皱长袖衬衫、深酒红高腰长裙。不得年轻化男主，不得替换女主的脸，不得改变衣领、纽扣和服装，不添加首饰。
空间锁定：两人在同一间植物书房的同一小块地面近距离面对面，相距约60至70厘米，始终男左女右，眼神相向。不是分别站在走廊两端，也不是两张单人照片拼接。两人之间没有门柱、桌子或隔断。相机留在原定妆图所示的对话轴线同一侧，可改变景别而不能反转人物左右。
风格锁定：「混合宅邸·精致暗彩」。墨绿墙、深胡桃木护墙和背景实体雕花门套形成一至两层纵深，局部细花纹与素面交替；右侧植物纹彩绘玻璃窗投来柔和日光，深木压住奶油白衣亮部，少量梅紫酒红织物。与参考保持同一时刻同一光向，暗部可辨，肤色真实，白衣不过曝。人物面部落在安静的背景上，门框均在身后，不穿过头或身体。环境现有植物标本册、玻璃植物箱、黄铜灯只可在取景包含时出现，不增添新道具。真实材质与细小肤纹，高质量电影剧照，不额外添加胶片颗粒、扫描痕迹或噪点。
画面里只能出现此男女两人。无字幕、无对话气泡、无文字、无编号、无水印、无手机界面。没有拥抱、亲吻、触碰对方、指戳脸或肢体冲突。表演应真实鲜活，不是广告摆拍；除非本镜头另有描述，正在说话的人嘴略张、听话者闭嘴，不让两人同时夸张张嘴。双手自然，避免遮脸、交叉或多指。
本镜头：第3—4轮对白的冲突高点。镜头比参考推进为较紧的腰胸以上双人中近景，略低于眼平但不夸张，双方脸都清楚可读，不把女主裁成后脑。男左女右保持同样近距离。男主被女方反驳后正脱口说“太荒唐了”的瞬间：眉毛明显压低，额头和鼻根有真实细纹，下颌紧绷，嘴自然张开正在发出有力量的短句，身体从腰以上前倾少许，一只手掌在胸腰之间斜向外打开、手指自然分开，另一只手在画外，不指人、不碰人。这是愤怒与不敢相信的组合，不是滑稽吼叫或咬牙露齿。女主站直、重心没有后退，脸上先受刺痛再转为审视的神情：眼睛轻微睁大，眉心拢起、嘴唇闭紧，下巴微抬，双手已从攥紧变为自然垂下；她坚定看着他，准备冷静反问。强调男方前倾形成的斜线对比女方垂直稳定的身体。两人的表情与第一张明显不同，环境侧光连续。

subject_definitions:
<Subject 1> is the adult man on the left in <Picture 1>, with tousled dark hair, light stubble, an ivory shirt with rolled sleeves, beige trousers and a dark belt.
<Subject 2> is the adult woman on the right in <Picture 1>, with a low dark bun, an ivory high-neck long-sleeved blouse and a muted burgundy long skirt.
<Subject 3> is the mansion study in <Picture 1>, with deep green walls, carved dark walnut, layered doorways, stained-glass window light and a wooden desk.
<Picture 1> is the composition and performance-blocking reference for [Shot 1], placing both characters close together in one frame.
<Audio 1> is the male voice-identity reference for <Subject 1> (S1).
<Audio 2> is the female voice-identity reference for <Subject 2> (S2).

summary:
[reference generation + audio reference] The target video is an 11-second escalation in which the man's explosive accusation meets the woman's angry challenge, using <Picture 1> for <Subject 1>, <Subject 2>, <Subject 3> and their close two-shot; <Audio 1> and <Audio 2> provide the separate male and female voice identities.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the man's facial identity, hair, apparent age and clothing throughout.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the woman's facial identity, hair, apparent age and clothing throughout.
<Subject 3> (appears in [Shot 1]): fully_preserved - retain the mansion study, materials, palette and window-light direction.
<Picture 1> ([Shot 1] composition reference): fully_preserved - preserve the close two-person arrangement and viewpoint while allowing the specified acting.
<Audio 1>: reference - retain the male speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.
<Audio 2>: reference - retain the female speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.

detailed_description:
The target video is a 16:9 realistic dramatic film scene with deep green walls, dark carved wood, subdued burgundy fabric and ivory clothing. Soft directional window light shapes natural faces, while the intense argument remains the focus; there are no subtitles, captions, logos or additional people.
Both voices remain recognizable as their respective audio references. At the specified peaks, anger produces a loud projected shout with a raw, hoarse, grainy edge and a brief strained vocal break; pitch and roughness may rise naturally without changing the speaker's identity. This is an audible shouted performance with clear syllables, not a whisper, a breathy hiss, a calmly recited sentence or merely a louder recording. Preserve clean microphone sound without digital distortion.
[Shot 1] Maintain the close eye-level medium two-shot planned by <Picture 1> in <Subject 3>. <Subject 1> is on the left and <Subject 2> on the right, within conversational reach and with both faces unobstructed. Their anger is already active at the opening: tightened jaws, sustained eye contact and compressed breathing. Use a nearly still camera, with no cut or dramatic camera move competing with the argument.
<Subject 1> (S1), in the same male voice identity referenced from <Audio 1>, starts almost immediately. Furious that she has rejected the role he assigns her, he SHOUTS THE ENTIRE OPENING EXCLAMATION with forceful chest projection and a raw, hoarse, cracking edge: <d>[Chinese] 太荒唐了，</d> The vocal attack begins on the first syllable. His jaw drops, his brows press inward, and one sharp opening of his right hand accompanies the burst.
After only a short catch of breath, about a fifth of a second, <Subject 1> (S1) drives the accusation straight at her in a sustained angry shout: <d>[Chinese] 就这样把你最神圣的责任丢下不管了呀。</d> The connecting words move briskly, while the phrase about her most sacred responsibility carries hard, biting emphasis. The final accusation about abandoning it builds to another hoarse, chest-driven peak. Keep the ending particle fully voiced and intelligible; the voice releases only after that syllable finishes. His torso pitches slightly toward her with the renewed vocal force, then his raised hand lowers as the sentence ends. Preserve his recognizable male timbre while anger temporarily raises his pitch and roughens the voice. The two dialogue blocks form one continuous accusation in approximately the first five and a half seconds. Speak every word once and in order, using a short breath between the blocks, without a long pause, added words or a wordless scream. He closes his mouth and holds his glare while she answers.
After a brief turn-taking breath, <Subject 2> (S2), using the same female voice identity from <Audio 2>, takes one small half-step toward him, stopping close without touching him. Her eyes narrow in disbelief, her brows pull together and her chin thrusts slightly forward. She delivers a short, firm command demanding his answer: <d>[Chinese] 你说，</d>
She then PAUSES FOR APPROXIMATELY HALF A SECOND. This is a clearly audible silence inside her own turn: her mouth stops moving, her lips come gently together, and she holds a hard, expectant stare at him. He stays silent. Preserve this deliberate pause after the opening two syllables, separate from the brief gap between the two speakers.
After that half-second hold, the same <Subject 2> (S2) continues in the same voice identity from <Audio 2> with an openly angry, rising challenge: <d>[Chinese] 什么是我最神圣的责任？</d> Her voice grows loud and raspy on the phrase about her most sacred responsibility, pressing him for an answer. Her lips and jaw articulate the final two syllables fully while the questioning contour remains forceful. The two blocks are one woman's continuous speaking turn with one deliberate internal pause. Speak the words in order exactly once. The man stiffens under her challenge and listens without mouthing her words.
Let the man's accusation occupy approximately the first five and a half seconds. The woman then speaks her two-syllable opening, holds the specified half-second silence, and delivers the complete remaining question, aiming to finish around nine and a half to ten seconds. These are pacing targets, with the complete words and the deliberate internal pause taking priority. Keep the turn-taking gap between the two speakers short; retain the separate half-second pause inside the woman's line. Leave at least the final second of the eleven-second shot for the complete final syllable to settle, a natural breath and their unwavering stare. Neither character smiles, adds a word or begins another sentence.

overall_soundscape:
Low indoor room tone, light clothing movement and natural breaths remain underneath the dialogue. The projected voices dominate the sound field with a short, natural indoor reflection. No competing voices, laughter or sound effects cover the final syllables.

non_diegetic_music:
N/A

用途：identity-preserve，电影对白二创的独立分镜照片。生成一张16:9横屏写实电影静帧，单张完整画面，不是拼图。参考图是本项目已确定的双人定妆照，作为唯一的人物身份、服装、空间和光线连续性锚点。
锁定人物：恰好两名成年人。画面左侧是参考中成熟东亚男性，保留其具体脸型、鼻梁、略显疲惫的眼睛、蓬松微乱短黑发与淡胡茬，象牙白卷袖亚麻衬衫、米色长裤、深色腰带。画面右侧是参考中的同一成年东亚女性，保留其具体五官、低盘髻和鬓边碎发，奶油白闭合立领褶皱长袖衬衫、深酒红高腰长裙。不得年轻化男主，不得替换女主的脸，不得改变衣领、纽扣和服装，不添加首饰。
空间锁定：两人在同一间植物书房的同一小块地面近距离面对面，相距约60至70厘米，始终男左女右，眼神相向。不是分别站在走廊两端，也不是两张单人照片拼接。两人之间没有门柱、桌子或隔断。相机留在原定妆图所示的对话轴线同一侧，可改变景别而不能反转人物左右。
风格锁定：「混合宅邸·精致暗彩」。墨绿墙、深胡桃木护墙和背景实体雕花门套形成一至两层纵深，局部细花纹与素面交替；右侧植物纹彩绘玻璃窗投来柔和日光，深木压住奶油白衣亮部，少量梅紫酒红织物。与参考保持同一时刻同一光向，暗部可辨，肤色真实，白衣不过曝。人物面部落在安静的背景上，门框均在身后，不穿过头或身体。环境现有植物标本册、玻璃植物箱、黄铜灯只可在取景包含时出现，不增添新道具。真实材质与细小肤纹，高质量电影剧照，不额外添加胶片颗粒、扫描痕迹或噪点。
画面里只能出现此男女两人。无字幕、无对话气泡、无文字、无编号、无水印、无手机界面。没有拥抱、亲吻、触碰对方、指戳脸或肢体冲突。表演应真实鲜活，不是广告摆拍；除非本镜头另有描述，正在说话的人嘴略张、听话者闭嘴，不让两人同时夸张张嘴。双手自然，避免遮脸、交叉或多指。
本镜头：第5—6轮对白中女主说到“可是”的情绪转折。改变机位但保持对话轴线同一侧，采用偏女主的双人中近景：女主右侧占约55%注意力，头部到腰部可见，面孔相对于镜头更接近四分之三正面；男主左侧占35%，仍能清楚看到他的脸、肩胸，不能只留一个后脑或模糊肩膀。人物在房间里的站位与60至70厘米距离不变，构图偏移来自镜头，不靠把人物挪远。女主正在认真开口说话，嘴唇自然打开，眼睛润泽且眼尾略红，但不流泪；眉心的痛感未消，下巴已稳住，胸口因深吸气略抬，肩膀打开，她的一只手在自己的腰胸下方微微摊开，掌心向上，手势小而诚恳，表示她也承担对自己的责任，绝不是教训式指点。男方原先列举的手势已停住，嘴闭紧，眉间不耐烦与一瞬困惑并存，直视女方；两只手不要和女主的手交叠。表现由伤心到争取尊严的活跃情绪。背景的木框和绿墙柔和退后，右侧花窗光照亮女主脸颊与眼睛，男方暗面仍有细节，景深足够使两张脸都清楚。

subject_definitions:
<Subject 1> is the adult man on the left in <Picture 1>, with tousled dark hair, light stubble, an ivory shirt with rolled sleeves, beige trousers and a dark belt.
<Subject 2> is the adult woman on the right in <Picture 1>, with a low dark bun, an ivory high-neck long-sleeved blouse and a muted burgundy long skirt.
<Subject 3> is the mansion study in <Picture 1>, with deep green walls, carved dark walnut, layered doorways, stained-glass window light and a wooden desk.
<Picture 1> is the composition and performance-blocking reference for [Shot 1], placing both characters close together in one frame.
<Audio 1> is the male voice-identity reference for <Subject 1> (S1).
<Audio 2> is the female voice-identity reference for <Subject 2> (S2).

summary:
[reference generation + audio reference] The target video is a 10-second argument in which the man insists on family obligations and the woman forcefully counters with an equally sacred responsibility, using <Picture 1> for <Subject 1>, <Subject 2>, <Subject 3> and their close two-shot; <Audio 1> and <Audio 2> provide the separate male and female voice identities.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the man's facial identity, hair, apparent age and clothing throughout.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the woman's facial identity, hair, apparent age and clothing throughout.
<Subject 3> (appears in [Shot 1]): fully_preserved - retain the mansion study, materials, palette and window-light direction.
<Picture 1> ([Shot 1] composition reference): fully_preserved - preserve the close two-person arrangement and viewpoint while allowing the specified acting.
<Audio 1>: reference - retain the male speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.
<Audio 2>: reference - retain the female speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.

detailed_description:
The target video is a 16:9 realistic dramatic film scene with deep green walls, dark carved wood, subdued burgundy fabric and ivory clothing. Soft directional window light shapes natural faces, while the intense argument remains the focus; there are no subtitles, captions, logos or additional people.
Both voices remain recognizable as their respective audio references. At the specified peaks, anger produces a loud projected shout with a raw, hoarse, grainy edge and a brief strained vocal break; pitch and roughness may rise naturally without changing the speaker's identity. This is an audible shouted performance with clear syllables, not a whisper, a breathy hiss, a calmly recited sentence or merely a louder recording. Preserve clean microphone sound without digital distortion.
[Shot 1] Use <Picture 1> as the composition reference for a steady medium two-shot in <Subject 3>. <Subject 1> stands on the left and <Subject 2> on the right, close enough that each small reaction is visible beside the speaking face. Preserve their identities, clothing, doorway depth and side lighting. Begin in the ongoing argument, without a neutral introductory pose or a long silent lead-in.
<Subject 1> (S1), using the male voice identity of <Audio 1>, lifts one open palm between them. He believes the answer should be unquestionable and resents her challenge. With a hard stare and rigid jaw, he fires the first question in a raised, biting, impatient voice: <d>[Chinese] 这还用我说吗？</d>
After only a brief catch of breath, <Subject 1> (S1) SHOUTS THE ENTIRE ANSWER with forceful chest projection and a raw, hoarse, cracking edge: <d>[Chinese] 当然是你的丈夫和你的孩子。</d> The burst begins on the word meaning of course. Drive the clause forward and give the words meaning husband and children two distinct vocal peaks. Both roles are shouted clearly, with rough anger in the voice; the second peak remains forceful through its final syllable. His jaw opens wider at those accents and his palm makes one firm downward beat on the final role. Keep his recognizable male timbre through the rise in pitch and temporary rasp. He is imposing his answer on her, with no friendly smile or softened finish. The two dialogue blocks occupy approximately the first four seconds as one rapid, forceful turn. Keep the brief connecting words moving and preserve the complete final syllable. Say each block once, in order, with no added words, repeated roles or long silence. He closes his mouth after the line and listens while she answers.
<Subject 2> (S2), in the female voice identity from <Audio 2>, responds after a brief gap. Her initial acknowledgement is sharp and confrontational, not submissive agreement. On the turn into her counterargument, she lifts her chin and brings one open hand to her upper chest. Her voice rises into a forceful, hoarse assertion on the phrase meaning equally sacred, then carries that pressure through the word meaning responsibility. She says exactly once: <d>[Chinese] 没错，可是我还有同样神圣的责任。</d> Her mouth opens decisively on the stressed syllables, her breath drives the statement forward and her gaze does not leave his face. The phrase has a rough, anger-strained edge, without sobbing, a smile or a soft reassuring finish. The man lowers his palm and watches her, visibly frustrated at being challenged.
Aim for the man's complete line in roughly the first four seconds and the woman's complete response by about eight and a half seconds. Keep the turn-taking pause short, preserving a natural speaking rhythm rather than squeezing the final word. Let the last syllable of her sentence sound fully before her mouth closes. During the remaining second and a half, hold her hand near her chest and their tense eye contact, with a short natural exhale and room tone. Keep the framing and lighting stable through the final frame.

overall_soundscape:
Low indoor room tone, light clothing movement and natural breaths remain underneath the dialogue. The projected voices dominate the sound field with a short, natural indoor reflection. No competing voices, laughter or sound effects cover the final syllables.

non_diegetic_music:
N/A

用途：identity-preserve，电影对白二创的独立分镜照片。生成一张16:9横屏写实电影静帧，单张完整画面，不是拼图。参考图是本项目已确定的双人定妆照，作为唯一的人物身份、服装、空间和光线连续性锚点。
锁定人物：恰好两名成年人。画面左侧是参考中成熟东亚男性，保留其具体脸型、鼻梁、略显疲惫的眼睛、蓬松微乱短黑发与淡胡茬，象牙白卷袖亚麻衬衫、米色长裤、深色腰带。画面右侧是参考中的同一成年东亚女性，保留其具体五官、低盘髻和鬓边碎发，奶油白闭合立领褶皱长袖衬衫、深酒红高腰长裙。不得年轻化男主，不得替换女主的脸，不得改变衣领、纽扣和服装，不添加首饰。
空间锁定：两人在同一间植物书房的同一小块地面近距离面对面，相距约60至70厘米，始终男左女右，眼神相向。不是分别站在走廊两端，也不是两张单人照片拼接。两人之间没有门柱、桌子或隔断。相机留在原定妆图所示的对话轴线同一侧，可改变景别而不能反转人物左右。
风格锁定：「混合宅邸·精致暗彩」。墨绿墙、深胡桃木护墙和背景实体雕花门套形成一至两层纵深，局部细花纹与素面交替；右侧植物纹彩绘玻璃窗投来柔和日光，深木压住奶油白衣亮部，少量梅紫酒红织物。与参考保持同一时刻同一光向，暗部可辨，肤色真实，白衣不过曝。人物面部落在安静的背景上，门框均在身后，不穿过头或身体。环境现有植物标本册、玻璃植物箱、黄铜灯只可在取景包含时出现，不增添新道具。真实材质与细小肤纹，高质量电影剧照，不额外添加胶片颗粒、扫描痕迹或噪点。
画面里只能出现此男女两人。无字幕、无对话气泡、无文字、无编号、无水印、无手机界面。没有拥抱、亲吻、触碰对方、指戳脸或肢体冲突。表演应真实鲜活，不是广告摆拍；除非本镜头另有描述，正在说话的人嘴略张、听话者闭嘴，不让两人同时夸张张嘴。双手自然，避免遮脸、交叉或多指。
本镜头：第7—8轮对白结尾，女方完整说完“我说的是我对我自己的责任”之后的对视停留。采用四张中最紧的双人胸像近景，保持两张脸完整可辨，男在左，女在右；女主略靠画面前方但两人实际距离仍约60至70厘米。相机沿原轴线同一侧轻移，让女主是四分之三正面、眼睛朝画面左侧看男主，男方也保留三分之二侧脸，绝不正对镜头看观众。女主完成宣告后自然合上嘴唇，眼里有水光、眉眼含着受伤和释然但姿态坚定，下巴水平、肩膀落稳，目光不闪躲；不是微笑，不是胜利得意，不是愤怒尖叫。男主不再保持上张的硬气：眉间稍松、眼神短暂失去把握，嘴微微松开但不在说话，身体从前倾恢复直立，原先抬起的手完全放下到取景外。他被这句话击中，想反驳却说不出来。两人都不哭泣、不新增动作。用脸部很小但清晰的不同变化做情绪结尾。背景保留虚化但可识别的深木门套、墨绿墙和右侧窗光，浅景深不能把任何一张脸虚掉。皮肤细节真实，安静而有压力。

subject_definitions:
<Subject 1> is the adult man on the left in <Picture 1>, with tousled dark hair, light stubble, an ivory shirt with rolled sleeves, beige trousers and a dark belt.
<Subject 2> is the adult woman on the right in <Picture 1>, with a low dark bun, an ivory high-neck long-sleeved blouse and a muted burgundy long skirt.
<Subject 3> is the mansion study in <Picture 1>, with deep green walls, carved dark walnut, layered doorways, stained-glass window light and a wooden desk.
<Picture 1> is the composition and performance-blocking reference for [Shot 1], placing both characters close together in one frame.
<Audio 1> is the male voice-identity reference for <Subject 1> (S1).
<Audio 2> is the female voice-identity reference for <Subject 2> (S2).

summary:
[reference generation + audio reference] The target video is an 11-second climax in which the man's dismissive outburst is answered by the woman's fully shouted declaration of responsibility to herself, using <Picture 1> for <Subject 1>, <Subject 2>, <Subject 3> and their close two-shot; <Audio 1> and <Audio 2> provide the separate male and female voice identities.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the man's facial identity, hair, apparent age and clothing throughout.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the woman's facial identity, hair, apparent age and clothing throughout.
<Subject 3> (appears in [Shot 1]): fully_preserved - retain the mansion study, materials, palette and window-light direction.
<Picture 1> ([Shot 1] composition reference): fully_preserved - preserve the close two-person arrangement and viewpoint while allowing the specified acting.
<Audio 1>: reference - retain the male speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.
<Audio 2>: reference - retain the female speaker's recognizable timbre, accent and resonance; generate a new performance following the target dialogue and anger cues. Do not copy the source waveform, pauses or restrained delivery.

detailed_description:
The target video is a 16:9 realistic dramatic film scene with deep green walls, dark carved wood, subdued burgundy fabric and ivory clothing. Soft directional window light shapes natural faces, while the intense argument remains the focus; there are no subtitles, captions, logos or additional people.
Both voices remain recognizable as their respective audio references. At the specified peaks, anger produces a loud projected shout with a raw, hoarse, grainy edge and a brief strained vocal break; pitch and roughness may rise naturally without changing the speaker's identity. This is an audible shouted performance with clear syllables, not a whisper, a breathy hiss, a calmly recited sentence or merely a louder recording. Preserve clean microphone sound without digital distortion.
[Shot 1] Hold the close medium two-shot established by <Picture 1> in <Subject 3>, with <Subject 1> on the left and <Subject 2> on the right. Keep both faces large enough for their eye and mouth movements to remain legible. The argument has reached its climax. There is no calm reset, opening pause, camera cut or change of setting.
<Subject 1> (S1), using the male voice identity from <Audio 1>, loses his patience at her insistence on her own responsibility. He SHOUTS THE ENTIRE OPENING DENIAL from its first syllable, with forceful chest projection and a raw, hoarse, cracking edge: <d>[Chinese] 没有的事，</d> His jaw drops, his brows press inward and one sharp outward sweep of his hand lands with the explosive rejection.
He takes only a short catch of breath, about a fifth of a second, then <Subject 1> (S1) SHOUTS THE ENTIRE FOLLOWING QUESTION directly at her: <d>[Chinese] 你还有什么责任啊你？</d> The rough, angry projection continues from the first word, building again on the challenge about what responsibility she could possibly have. Keep the last two syllables fully audible with a forceful questioning contour. His frustration makes the pitch jump and the voice briefly crack, while his recognizable male timbre remains intact. His torso pitches slightly toward her at the vocal peak. After the final syllable has sounded completely, he closes his mouth and rotates his shoulders a quarter-turn away in dismissal, keeping his face readable and his body inside the frame. The denial and question form one continuous angry turn in roughly the first three and a half seconds. Speak each block once, in order, with no added words, long internal pauses or wordless scream. His voice has stopped before she begins her declaration.
After a short, charged breath, <Subject 2> (S2), in the same female voice identity from <Audio 2>, plants her stance and raises one hand firmly against her own upper chest. She delivers her ENTIRE final sentence as a loud, sustained, raw and hoarse angry shout, from the opening syllable to the very last syllable. She says exactly once: <d>[Chinese] 我说的是我对我自己的责任。</d> Begin with an emphatic shouted attack; drive through the middle without falling back to calm or quiet speech. Place the greatest force on the reference to herself and on the final word meaning responsibility. Allow an anger-strained, tearing rasp and a brief break in the voice, while keeping every Chinese syllable clearly audible. Her eyebrows pull inward, her eyes remain fierce and hurt, her mouth opens fully at the accented words, and her chest and shoulders follow one powerful exhale. Do not replace any syllable with a wordless scream, whisper or sob. Carry the last syllable, ren, all the way through its complete sound before the breath releases. This is a shouted assertion of dignity, not a soft closing reflection.
The man arrests his turn and looks back toward her as the declaration lands, keeping silent. Aim for his line in the first three and a half seconds and her complete shouted sentence to finish by about eight and a half to nine seconds. Reserve the final two seconds for her lips to close, a naturally audible exhale and the tense reaction between them. Her anger remains visible after the shout. Do not add dialogue, repeat the last word, introduce a smile, or fade out while she is still vocalizing.

overall_soundscape:
Low indoor room tone, light clothing movement and natural breaths remain underneath the dialogue. The projected voices dominate the sound field with a short, natural indoor reflection. No competing voices, laughter or sound effects cover the final syllables.

non_diegetic_music:
N/A