《天宫幻梦》

写实东方神话电影静帧，16:9。云海之上的雕花通天长阶沿画面纵深延伸，一位成年东方女子居中正面全身朝镜头缓步上阶，身形修长、步态端正，一手轻提裙摆。她梳高盘发与垂落长发，佩戴层叠金白花冠、少量红珠和垂珠耳饰，身穿红、青绿与金色织锦交领长袍，米白金纹内裙与半透明红纱披帛随高空微风向后飘。晨昏暖光从侧后方照亮云层、衣料金线和石阶纹样，真实皮肤、发丝和织物细节，电影级自然景深，不要多余人物、文字、水印、畸形手指或夸张塑料皮肤。

写实东方神话电影人像，16:9，中近景正面构图。同一位成年东方女子站在云海通天长阶上，身体和脸朝向镜头，双手自然整理腰前衣料，神情安静克制。固定同一张自然鹅蛋脸、棕色杏眼、真实肤质与黑色高盘发，层叠金白花冠带少量红珠，冠上长珠链与较短花形珠滴耳饰清楚；红、青绿、金色刺绣长袍和米白金纹交领保持一致，半透明红纱向画面右侧飘动。背景石阶和云海柔化但方向可辨，暖金侧光勾勒发丝、珠饰和锦缎，不改变人物身份、服装领口或头饰结构。

超写实电影微距特写，16:9，只呈现同一位女子的一只棕色眼睛及周围真实皮肤。睫毛根根清楚，眼睑、泪阜、细小毛孔与自然高光真实可见；湿润角膜和虹膜内部映出一条通往云端宫阙的金色长阶，宫殿、云海和夕照沿瞳孔透视收束，像真实光学倒影与梦境入口融合。保持原眼睛方向，不镜像、不换眼，云宫占据虹膜主体但仍保留虹膜纹理、角膜反光和眼球曲率。浅景深、柔暖侧光、无文字无水印，不要额外眼睛、卡通虹膜、机械眼或不自然发光。

写实东方神话电影静帧，16:9。成年东方女子背对镜头站在云海之上的通天长阶中央，完整全身与拖地裙摆清楚可见，视线朝向远处悬于云山上的宏伟宫阙。黑色长发从高盘发垂至后背，金白花冠轮廓清晰；红金织锦长袍、米白金纹内裙和半透明红纱披帛在高空风中向画面左侧舒展。雕花中轴与两侧石阶形成强烈引导线，夕阳从右侧穿过云层，照亮宫殿檐角、衣料金线和云海边缘。空间宏大、材质真实、自然电影色彩，不出现正面脸、第二人物、文字或漂浮建筑。

超宽幅东方神话电影大远景，16:9。镜头高位后撤，完整展开一条从前景穿越云海直达远端宫阙的漫长雕花天阶，成年华服女子缩小为长阶左侧的红金色视觉锚点，朝宫殿方向站立。宫阙位于云山之上，长阶、云海和天空形成清晰三层纵深；右侧夕阳暖光横扫云层与石栏，远处青灰云影保持冷暖平衡。女子服装仍是红金织锦与透明红纱，比例自然，不放大人物、不增加同行者。真实航拍电影质感，建筑透视稳定，避免插画化、过度雾化、文字和水印。

写实东方神话电影中景，16:9。同一位成年东方女子在云端石阶上止步，镜头从她右后侧拍摄，身体仍朝阶梯前方，头部保持侧向尚未回眸。黑色长发垂至背部，金白花冠与侧面珠饰结构稳定；红、青绿、金色织锦长袍和米白内裙保持原样，半透明红纱被高空风拉向画面右侧。左侧石阶和雕花栏杆向云海深处延伸，暖金夕光勾勒面部侧缘、发丝与衣料纹理。自然成年比例、真实皮肤和布料重量，不要正面凝视、第二人物、突然换装或头饰漂移。

写实电影侧脸近景，16:9。同一位成年东方女子站在云端长阶，身体背向镜头方向，头部只转到清晰左侧脸，目光仍望向画外，处于完成回眸之前的瞬间。固定自然鹅蛋脸、棕色杏眼、高盘发、后垂长发和层叠金白花冠，耳侧长珠链轻微下垂；红金织锦肩背与米白领口只露出合理局部。云海、石阶与夕阳成为柔和浅景深背景，暖光描出鼻梁、睫毛和颈侧轮廓。表情冷静克制，真实毛孔和发丝，不要提前看镜头、夸张笑容、变脸或多余首饰。

写实东方神话电影回眸近景，16:9。承接上一镜，同一位成年东方女子从肩后完成一次小幅回眸，脸转向镜头，棕色杏眼平静凝视，嘴唇自然放松。保持同一自然脸型、真实肤质、黑色高盘发和后垂长发，层叠金白花冠、少量红珠、冠链珠滴与花形耳坠位置准确；红金织锦肩部和米白金纹领口细节清楚。云海与石阶在背景中柔焦，暖金侧逆光照亮眼睛、发丝和金属珠饰。构图停在胸肩以上，不冲成眼部特写，不要磨皮、换脸、笑容或饰品变形。

写实东方神话电影低机位特写，16:9。镜头贴近坚硬的灰色湿润石台，只见成年女子象牙白与浅金刺绣窄裙的下摆自然垂落在画面上部，真实布料褶皱与花纹清晰；下方浅薄水光完整映出同一裙摆的倒影，石板接缝、冷暖天光和远处巨月圆环以虚化层次出现。倒影必须与实体裙摆和机位光学对应，不是第二个人，也不是深水池。材质写实、水平线稳定、无飞溅，无脚部畸形、画面倒置、文字或水印。

写实东方神话电影大远景，16:9。冷灰石台悬于云海，一株横向伸展的古松从画面左侧压入，巨大的完整圆环与苍白月轮位于右后方，同一位象牙白金绣华服女子以小比例站在环下石台中央偏左。人物身形修长、窄裙自然垂地，姿态安静，周围云山与晨昏天空层次清楚。古松、圆环、巨月和平台透视稳定，月亮不膨胀；冷蓝云影配少量暖金侧光，真实镜头空气感，不要插画化、额外人物、漂浮松枝或建筑变形。

写实东方神话电影全身像，16:9。同一位成年东方女子站在湿润灰石台上，完整修长身形与窄长裙摆清楚，双手自然交叠于腰前，神情清冷忧郁、不笑。固定自然脸型与黑色高盘发，金白小花饰带少量红珠，垂珠耳饰，象牙白与浅金花卉刺绣的露肩长裙及薄披帛保持一致。古松从左上方伸入，巨月与圆环在右后方，云海和远山形成深景，石面有克制反光。自然肤质和布料重量，避免红油肤色、宽大膨胀裙型、第二人物或松枝穿脸。

写实电影半身人像，16:9。同一位成年东方女子在巨月古松前低眸，脸略向画面右下方，双手在腰前轻拢宽袖，动作克制。她保持自然鹅蛋脸、黑色高盘发、金白花饰与少量红珠、细长垂珠耳饰，象牙白浅金刺绣露肩衣裙和薄披帛不变。左侧古松、右后巨月圆环与云海位置稳定，浅景深仍可辨识环境；柔冷月光为主，少量暖光擦过脸颊和衣料金线。保留毛孔、发丝与手部结构，不要笑、抬眼、过度磨皮或饰品漂移。

写实东方电影正面近景，16:9。承接低眸半身，同一位成年东方女子缓缓抬头抬眼，目光短暂落向镜头，表情清冷、忧郁而克制，唇角不笑。保持同一自然脸型、真实肤质、黑色高盘发、金白红珠花饰和垂珠耳饰，象牙白浅金刺绣领口只露出合理范围。左侧古松与右后巨月在浅景深中形成稳定背景，冠顶留在画幅内，柔冷光照亮眼睛与发丝，少量暖边光勾勒脸颊。不要继续推进成眼部特写，不换脸、不红油磨皮、不添加夸张表情。

写实东方神话电影大远景，16:9。云海之上的浅水石岸横贯前景，晶角白鹿在画面左侧低头饮水，同一位深紫银绣华服女子站在右侧干岸，二者倒影落入平静水面。远处层叠云宫在夕阳中漂浮于云海，建筑、水面、峭壁与暖金天空形成宏大纵深。女子比例较小，窄裙垂直，白鹿四肢、晶体鹿角与倒影结构准确；人物右、鹿左的关系清楚。真实电影空气感和冷暖平衡，不要多余鹿、人物入水、建筑漂移、卡通童话质感或水印。

写实东方神话电影全身画面，16:9。晶角白鹿位于左侧浅水中低头饮水，成年东方女子站在右侧干燥石岸，完整全身和修长窄裙清楚，双手自然收于腰前。女子穿深紫偏蓝的细银绣露肩长袍，佩戴红金层叠花冠、珍珠垂链与红宝耳饰，保持自然成年比例和真实肤质。夕照云宫与水面倒影位于后方，白鹿晶角通透但结构稳定，水面只有细小涟漪。人物不入水，鹿不复制，服装不回到旧红衣，不出现畸形四肢或插画滤镜。

写实东方电影近景人像，16:9。同一位成年东方女子位于画面中央偏右，头部和棕色眼睛略向左下方望向画外白鹿，神情温柔克制，嘴唇自然。固定同一张自然脸、黑色高盘发、红金花冠、成串珍珠流苏与红宝耳饰；深紫银绣衣领仅露出合理局部。夕照云宫和云海在浅景深中柔化，右侧暖光勾勒皮肤、发丝与珠链，保持自然毛孔和细微色差。不要让白鹿角穿入脸部，不换脸、不抬头直视镜头、不添加微笑或旧红色服装。

写实东方神话电影正面近景，16:9。承接望鹿镜头，同一位成年东方女子将视线和头部小幅收回，正面安静凝视镜头，表情克制妩媚但不笑。保留自然鹅蛋脸、真实肤质、棕色杏眼、黑色高盘发、红金层叠花冠、珍珠流苏和红宝耳饰，深紫偏蓝细银绣服装保持连续。云宫、云海和夕阳成为柔和背景散景，暖侧光照亮眼睛与珠饰，冠顶完整。不要过度磨皮、油亮红肤、夸张嘴型、头饰变形或额外人物。

写实东方神话电影空场大景，16:9。巨大的白色云瀑从左侧和画面中央向下翻涌，云体具有厚度、纹理与自然阴影，不是发光液体；深色宫殿檐角与红柱从云瀑间显露，建筑牢固嵌在峭壁与云层之间。前景为低调深色石台，冷灰蓝天空与暖色宫柱形成克制对比。画面完全无人、无残留衣摆或人影，云瀑体积宏大但宫檐、柱子和地平线不变形。真实电影级空气透视，无文字、水印、卡通感或近镜头白雾遮挡。

写实东方神话电影正面半身人像，16:9。一位成年东方女子居中站在云瀑宫殿前，身体和脸正对镜头，双手自然交叠于腰前，神情安静庄重。固定自然脸型、真实冷白肤质、黑色中分半盘发、精致金冠和长金珠耳链；白色毛领覆盖双肩，内穿米白金纹胸衣，外层红青金色织锦宽袖层次清楚。左侧云瀑持续下落，右后宫殿红柱和暗色檐口稳定可辨，柔冷环境光配少量暖金侧光。不要第二人物、过度裸露、塑料皮肤、毛领变形或建筑漂移。

写实东方神话电影正面近景，16:9。承接半身镜头，同一位成年东方女子面向镜头，棕色眼睛柔和凝视，嘴唇自然放松，表情克制。保持同一自然脸型与真实肤质、黑色中分半盘发、金色冠饰、长金珠耳链、米白金纹领口和白色毛领；冠顶接近画幅上缘但完整可辨。巨型云瀑与宫殿红柱在浅景深背景中保持原方位，柔冷光照亮皮肤，金饰只有细小高光。不要继续冲眼、眨眼卖萌、露齿大笑、换脸、油亮磨皮或耳饰漂移。

subject_definitions:
<Subject 1> is ONE adult woman, primarily identified by <Picture 3>: the natural facial outline without jaw slimming, brown almond-shaped eyes, source nose and lip shapes, and natural skin texture; a high black updo with front wisps and long rear hair; a large layered gold-and-white floral crown with a few red beads and long bead-drop chains attached to the crown; and shorter flower-shaped pearl-drop earrings attached to the ears. Her visible cream-and-gold wrapped neckline and red, teal and gold embroidered robe draped to expose the upper arms follow that master. <Picture 4> assists side and back views of this SAME woman, not extra people or a panel layout. The master has appearance priority from opening to ending. Unseen lower garments keep the composition references' existing slender narrow silhouette.
<Subject 2> is the same long stairway above the cloud sea in <Picture 2>. The physical-left walking lane appears screen right in the frontal reverse view, with the central dragon-carved ramp on screen left. The heroine ascends toward the camera; behind her, the stairs descend toward the clouds.
<Picture 1> provides partial full-body composition and gait guidance only; it does not define the canonical face, jewellery, neckline or stair-height geometry.
<Picture 2> supplies the correct stair direction and medium-close composition; its face, crown, earrings and neckline are subordinate to <Subject 1>.

summary:
[reference generation] Generate one four-second frontal ascending shot of <Subject 1> in <Subject 2>, progressing from the body scale and gait intent of <Picture 1> toward the medium-close composition of <Picture 2>. Use the master appearance from <Picture 3> immediately, with <Picture 4> only assisting the same woman in other views; keep it unchanged during the approach. The two composition images guide framing and gait, while <Subject 1> defines the unchanged appearance.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the canonical face, skin, crown with its own long chains, shorter floral earrings, hair and visible wrapped neckline from <Picture 3> throughout; <Picture 4> provides same-person side/back guidance only. Keep the narrow lower-garment silhouette from composition guidance. Neither identity image supplies scenery, sheet background, panels, labels or additional women.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the single physical-left ascending lane and descending background stairway defined by <Picture 2>.
<Picture 1> ([Shot 1] full-body scale and gait): partially_preserved - retain full-body scale, placement and skirt-managing gait only; exclude its conflicting appearance and reversed tread-height geometry.
<Picture 2> ([Shot 1] appearance and close composition): partially_preserved - retain correct stair direction and medium-close composition; facial and upper-costume identity comes only from the canonical <Subject 1>.

detailed_description:
Live-action skin shows natural pores and subtle colour variation. Preserve normal slender adult proportions and weighted vertical robe folds, blue-grey cloud shadows, neutral-grey stone, warm side-edge light only, and restrained silk/metal highlights. Avoid an overall red-orange wash, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open at the full-body scale and placement planned by <Picture 1>, with <Subject 1> already the woman from <Picture 3>. Her layered gold-and-white floral crown with sparse red beads, crown-attached long chains, shorter floral pearl-drop earrings, high dark updo and cream-and-gold wrapped neckline are present immediately. The robe exposes her upper arms as in the master; its unseen lower part keeps the reference's slender narrow silhouette. Do not change from another starting face, crown or neckline. She is on the physical-left walking lane, which appears screen right in this frontal view; the central dragon-carved ramp is screen left.
From 00:00.000 to 00:00.700, establish the direction while her body prepares one small uphill step. The camera is ahead of her, near the upper end of the visible route, looking down the ascent. The staircase behind her drops away toward the cloud sea. Use <Picture 2> for this height relationship, not the apparently reversed tread geometry in <Picture 1>. Her posture is upright, with a relaxed gaze along her route and no facial performance.
From 00:00.700 to 00:03.400, she takes only one or two short ascending steps toward the camera. Each front foot lands on a higher tread before accepting body weight; her body rises by that actual step height. The supporting leg follows naturally without downhill sliding or hovering. Her fingers manage the front skirt lightly so the fabric clears the next tread. Keep the same hand and garment relationships as the frame tightens.
The camera initially retreats a little more slowly than her approach, then makes a restrained forward adjustment to reach the medium-close scale of <Picture 2>. Real steps and camera distance carry the shot from full body toward upper body. Keep the viewing axis and face proportions stable; do not replace her with a larger static portrait. The crown and neckline remain physically unchanged, and the earrings show only small motion caused by her steps.
World sunlight is on her bodily right, appearing screen left in the frontal reverse view. The established right-to-left world wind carries hair and trailing gauze screen right. Keep the grey treads continuous below her and the carved ramp on the correct side.
From 00:03.400 to 00:04.000, ease into a brief medium-close moment with her head and gaze steady. Her expression stays composed, with naturally relaxed lips. This short interval prepares the next eye push without adding a long photograph hold. The shot is continuous and completely silent.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, wind, ambient sound, or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman, primarily identified by <Picture 3>: the natural facial outline without jaw slimming, brown almond-shaped eyes, source nose and lip shapes, and natural skin texture; a high black updo with front wisps and long rear hair; a large layered gold-and-white floral crown with a few red beads and long bead-drop chains attached to the crown; and shorter flower-shaped pearl-drop earrings attached to the ears. Her visible cream-and-gold wrapped neckline and red, teal and gold embroidered robe draped to expose the upper arms follow that master. <Picture 4> assists side and back views of this SAME woman, not extra people or a panel layout. The master has appearance priority from opening to ending. Unseen lower garments keep the composition references' existing slender narrow silhouette.
<Subject 2> is the deliberately enlarged surreal cloud-palace scene occupying most of the iris region in <Picture 2>, including the long carved ascent and palace silhouette.
<Picture 1> is the opening medium-close composition for [Shot 1].
<Picture 2> guides macro framing and the chosen large eye-contained palace scene; the anatomical right eye itself follows <Subject 1>, with the tear caruncle at image right and no mirrored or swapped eye. The vision remains dominant rather than being obscured by a separate complete black pupil.

summary:
[reference generation] Generate a four-second continuous push from <Subject 1>'s medium-close view in <Picture 1> toward the selected macro composition of <Picture 2>. A short steady opening leads to a slow-then-fast approach, then deceleration so <Subject 2> remains legible across most of the iris. Preserve the surreal vision and photographic skin together. Character appearance follows <Picture 3> first, with <Picture 4> only assisting the same woman's other views; the preceding composition images do not override it.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the canonical face, skin, crown with its own long chains, shorter floral earrings, hair and visible wrapped neckline from <Picture 3> throughout; <Picture 4> provides same-person side/back guidance only. Keep the narrow lower-garment silhouette from composition guidance. Neither identity image supplies scenery, sheet background, panels, labels or additional women.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the large eye-contained palace scene rather than shrinking it to an ordinary tiny reflection.
<Picture 1> ([Shot 1] medium-close opening): partially_preserved - retain frontal medium-close framing and the head-settling pose; use the canonical face, crown, earrings and neckline.
<Picture 2> ([Shot 1] macro destination): partially_preserved - retain right-eye macro orientation and the large palace-in-iris scene; use the canonical eye contour and skin, without an obscuring enlarged black pupil.

detailed_description:
Live-action skin shows natural pores and subtle colour variation. Preserve normal slender adult proportions and weighted vertical robe folds, blue-grey cloud shadows, neutral-grey stone, warm side-edge light only, and restrained silk/metal highlights. Avoid an overall red-orange wash, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open in the medium-close frontal composition of <Picture 1>, showing the canonical <Subject 1> from <Picture 3> immediately. Preserve her brown almond eyes, source nose and lips, natural skin, layered floral crown with crown-attached chains, shorter floral pearl-drop earrings, high dark updo and cream-and-gold wrapped neckline with exposed upper arms. Her head is already settled after the short uphill approach. Keep the identity continuous as the frame narrows, with the upper cloud stairway still softly visible behind her.
From 00:00.000 to 00:00.600, hold the head and gaze briefly steady. Align the camera with her anatomical right eye, screen left in the frontal view. Keep the macro tear caruncle at image right; never mirror or switch eyes. Her lips stay relaxed without speech. World sunlight remains screen left in this frontal view, and any visible fine hair continues toward screen right under the same breeze.
From 00:00.600 to 00:01.300, begin a slow, precise forward push. Move through a closer face composition so that the eye's position, facial contour and light direction can be read. Crown and earrings simply leave the crop as distance closes; they do not change design. Maintain a natural photographic perspective instead of stretching the face into the lens.
From 00:01.300 to 00:02.700, accelerate toward the eye. The increasing approach speed creates the visual impact. Focus travels continuously from the face to the lashes and eye surface without a cut, exposure flash or blinking wipe. Keep skin texture, individual lashes, the moist lid edge and the white-of-eye boundary readable through the narrowing crop.
From 00:02.700 to 00:04.000, decelerate into the framing of <Picture 2>, keeping the same canonical right eye. <Subject 2>, the long carved staircase and cloud palace, becomes large and legible across most of the iris region. Honour this intentionally surreal image rather than reducing it to a small physically ordinary catchlight. Do not introduce a dominant complete black pupil that obscures the chosen palace scene.
The palace roofline and stair centreline keep their selected relationship as the approach slows. The outer lid, skin pores, lashes and scleral detail remain naturally photographed, giving the fantasy image a real human frame. Keep the macro stable enough to read its architecture during the final deceleration, without a separate long still-image hold. The heroine does not wink or close the eye. End ready for the next optical expansion, with one continuous shot and complete silence.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, wind, ambient sound, or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman, primarily identified by <Picture 3>: the natural facial outline without jaw slimming, brown almond-shaped eyes, source nose and lip shapes, and natural skin texture; a high black updo with front wisps and long rear hair; a large layered gold-and-white floral crown with a few red beads and long bead-drop chains attached to the crown; and shorter flower-shaped pearl-drop earrings attached to the ears. Her visible cream-and-gold wrapped neckline and red, teal and gold embroidered robe draped to expose the upper arms follow that master. <Picture 4> assists side and back views of this SAME woman, not extra people or a panel layout. The master has appearance priority from opening to ending. Unseen lower garments keep the composition references' existing slender narrow silhouette.
<Subject 2> is the same cloud palace and long carved stairway represented inside the eye in <Picture 1> and as full spatial scenery in <Picture 2>.
<Picture 1> is the opening macro-eye and scene-expansion reference for [Shot 1].
<Picture 2> is the complete rear-camera space and full-body placement reference for [Shot 2], reached through an optical matching hidden cut, not through the heroine materialising.

summary:
[reference generation] Generate a four-second optical transition from <Picture 1> to the complete rear-view world of <Picture 2>. In [Shot 1], <Subject 2> expands from the eye until its architecture fills the frame. At about 00:02.300, an aligned hidden cut switches to [Shot 2], where <Subject 1> is already present as a complete person on the stairs. Character appearance follows <Picture 3> first, with <Picture 4> only assisting the same woman's other views; the preceding composition images do not override it.

retention_analysis:
<Subject 1> (appears in [Shot 1] as the eye and [Shot 2] as the rear figure): fully_preserved - retain the canonical face, skin, crown with its own long chains, shorter floral earrings, hair and visible wrapped neckline from <Picture 3> throughout; <Picture 4> provides same-person side/back guidance only. Keep the narrow lower-garment silhouette from composition guidance. Neither identity image supplies scenery, sheet background, panels, labels or additional women.
<Subject 2> (appears in [Shot 1], [Shot 2]): fully_preserved - keep the palace silhouette, stair direction and cloud world consistent across the matching cut.
<Picture 1> ([Shot 1] macro opening): partially_preserved - retain macro framing and the large iris-contained scene during expansion; the surrounding eye and skin belong to the canonical <Subject 1>.
<Picture 2> ([Shot 2] rear composition): partially_preserved - retain complete body placement, planted stance, stair lane and established spatial depth; replace conflicting crown, earrings, hair and visible upper clothing with the canonical <Subject 1>.

detailed_description:
Live-action skin shows natural pores and subtle colour variation. Preserve normal slender adult proportions and weighted vertical robe folds, blue-grey cloud shadows, neutral-grey stone, warm side-edge light only, and restrained silk/metal highlights. Avoid an overall red-orange wash, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open on the canonical woman's right eye from <Picture 3>, using <Picture 1> for macro framing and the large fantasy vision. The large vision of <Subject 2> already occupies most of the iris, surrounded by natural photographed lids, lashes and skin. The palace roofline sits above the long carved ascent, providing two clear matching elements. Keep the chosen surreal image intact instead of restoring a dominant complete black pupil.
From 00:00.000 to 00:00.400, allow only a short readable eye moment. Then move forward into the palace image as a stylised optical expansion, not a journey through biological tissue. The architecture becomes larger while its stair centreline and roof remain coherent. The eye's outside boundary moves toward the edges of the crop as the scene gains the frame.
From 00:00.400 to 00:01.800, increase the expansion steadily. Let engraved stone, cloud layers and the palace silhouette acquire readable scene depth, while preserving their selected orientation. Do not generate a new person inside the iris or build a silhouette from clouds. The effect changes the visual plane of the palace image, not the heroine's body.
From 00:01.800 to 00:02.300, bring the palace and long staircase fully across the frame. Align the roofline and carved ascent with the corresponding world in <Picture 2>. A passing edge of bright cloud and the matched stone lines briefly cover the optical boundary, providing a natural mask for the planned cut.
[Shot 2] At 00:02.300, a hidden optical match cut switches to the complete rear-camera space of <Picture 2>. <Subject 1> is already fully present in its first image: a grounded adult figure on the physical-left walking lane, seen screen left beside the central carved ramp. Her layered floral crown, crown-attached chains, shorter floral earrings, high black updo, long rear hair and visible wrapped neckline follow <Picture 3>, with rear-view guidance from <Picture 4>. The existing narrow lower robe keeps normal scale, continuous edges and fabric weight. There is no fade-in, growth or emergence from the eye or cloud mask.
From 00:02.300 to 00:03.500, make a small rearward camera adjustment and settle the scene's perspective. Her feet stay planted while hair and trailing gauze continue the established breeze. Sunlight appears screen right from behind, and hair and fabric move screen left.
From 00:03.500 to 00:04.000, finish the settling motion without a long still hold. Keep the same palace ahead and one complete figure in the existing stair space, ready for the next backward reveal. Both shots remain entirely silent.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, wind, ambient sound, or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman, primarily identified by <Picture 3>: the natural facial outline without jaw slimming, brown almond-shaped eyes, source nose and lip shapes, and natural skin texture; a high black updo with front wisps and long rear hair; a large layered gold-and-white floral crown with a few red beads and long bead-drop chains attached to the crown; and shorter flower-shaped pearl-drop earrings attached to the ears. Her visible cream-and-gold wrapped neckline and red, teal and gold embroidered robe draped to expose the upper arms follow that master. <Picture 4> assists side and back views of this SAME woman, not extra people or a panel layout. The master has appearance priority from opening to ending. Unseen lower garments keep the composition references' existing slender narrow silhouette.
<Subject 2> is the same long ascending staircase, central dragon-carved ramp, palace and cloud sea in both images. The physical-left walking lane is screen left in this rear view.
<Picture 1> is the full-body rear composition and real step-planning reference for [Shot 1].
<Picture 2> is the complete far-wide composition approached by the continuous backward and upward camera movement in [Shot 1].

summary:
[reference generation] Generate one five-second rear-view reveal from <Picture 1> toward <Picture 2>. <Subject 1> makes one genuine small uphill step while the camera begins to withdraw, then the camera accelerates backward and rises moderately to expand <Subject 2>. Finish deceleration by about 00:03.800 and preserve the complete wide view until the end. Character appearance follows <Picture 3> first, with <Picture 4> only assisting the same woman's other views; the preceding composition images do not override it.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the canonical face, skin, crown with its own long chains, shorter floral earrings, hair and visible wrapped neckline from <Picture 3> throughout; <Picture 4> provides same-person side/back guidance only. Keep the narrow lower-garment silhouette from composition guidance. Neither identity image supplies scenery, sheet background, panels, labels or additional women.
<Subject 2> (appears in [Shot 1]): fully_preserved - reveal more of the existing stair-palace world without changing its layout or relocating the heroine.
<Picture 1> ([Shot 1] initial rear composition): partially_preserved - retain rear viewpoint, physical stair placement and the reveal composition; preserve the canonical head and upper-body appearance, with the existing slender lower robe and train.
<Picture 2> ([Shot 1] far-wide composition): partially_preserved - retain rear viewpoint, physical stair placement and the reveal composition; preserve the canonical head and upper-body appearance, with the existing slender lower robe and train.

detailed_description:
Live-action skin shows natural pores and subtle colour variation. Preserve normal slender adult proportions and weighted vertical robe folds, blue-grey cloud shadows, neutral-grey stone, warm side-edge light only, and restrained silk/metal highlights. Avoid an overall red-orange wash, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open behind <Subject 1> at the full-body scale of <Picture 1>. She occupies the same physical-left stair lane, seen screen left of the central dragon-carved ramp, with the palace ahead along the ascent. Her layered floral crown and crown chains, shorter floral earrings, high dark updo with long rear hair, wrapped neckline and exposed upper arms already follow <Picture 3>, aided by <Picture 4> for this rear angle. Keep the existing narrow lower robe and translucent train rather than inventing a new full outfit. Grey stone treads remain visible beneath her body, providing the scale for the coming step.
During the opening, she makes one small uphill step. The front foot reaches the next higher tread, lands firmly and begins accepting her weight; the pelvis and torso rise by that tread height. The person genuinely moves before the view expands. Keep it to one step, with natural balance and no further walking loop as the view expands.
From 00:00.500 to 00:01.500, start a controlled backward dolly and moderate upward crane while the step finishes its weight transfer. Maintain the established ascent axis and a steady focal length. Near stairs and the heroine shrink quickly with clear parallax, while the distant palace changes scale much less; do not uniformly shrink the whole image. Her robe gathers briefly with the step; its train stays continuous.
From 00:01.500 to 00:03.000, accelerate the retreat and rise enough to reveal the surrounding cloud sea. The camera movement now supplies the stronger spatial impact. Keep the same palace roof and terraces in place; the carved ramp remains one continuous route to them. The figure does not jump to the summit, cross the ramp or change lanes. Reveal the original staircase without inventing new treads.
From 00:03.000 to 00:03.800, progressively reduce both backward and upward speed as the framing approaches <Picture 2>. Let foreground stone, the small heroine and distant architecture settle into the complete far-wide composition. Avoid an abrupt zoom snap, rolling horizon or a new view inserted by a cut.
From 00:03.800 to 00:05.000, keep the camera fully steady on the broad world. The same world sun appears screen right behind her, and the fixed breeze carries hair and gauze screen left. Only subtle natural cloth and cloud motion continue; her supporting stance is settled. The palace and ascent remain readable for the following editorial cut. Do not fly from this wide ending directly onto her face. This generation contains one continuous silent shot.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, wind, ambient sound, or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman, primarily identified by <Picture 4>: the natural facial outline without jaw slimming, brown almond-shaped eyes, source nose and lip shapes, and natural skin texture; a high black updo with front wisps and long rear hair; a large layered gold-and-white floral crown with a few red beads and long bead-drop chains attached to the crown; and shorter flower-shaped pearl-drop earrings attached to the ears. Her visible cream-and-gold wrapped neckline and red, teal and gold embroidered robe draped to expose the upper arms follow that master. <Picture 5> assists side and back views of this SAME woman, not extra people or a panel layout. The master has appearance priority from opening to ending. Unseen lower garments keep the composition references' existing slender narrow silhouette.
<Subject 2> is the same physical-left stair lane and cloud world seen around her in the three images, with unchanged world sunlight and wind.
<Picture 1> is the side-rear medium composition of an already-stopped person, selected after a separate editorial cut from the previous far-wide shot.
<Picture 2> is the closer pre-turn composition reached by a short camera approach.
<Picture 3> is the close backward-gaze composition reached through one physical head turn. Facial identity and the relaxed lip shape follow <Subject 1>, without added smiling or lip-corner lift.

summary:
[reference generation] Generate a four-second side-rear portrait shot beginning with <Subject 1> already stopped in the composition of <Picture 1>. Advance toward <Picture 2>, let her turn back once, and settle briefly into the gaze of <Picture 3>. This unit starts after an editorial cut; it does not continue the previous far-wide camera flight. Character appearance follows <Picture 4> first, with <Picture 5> only assisting the same woman's other views; the preceding composition images do not override it.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the canonical face, skin, crown with its own long chains, shorter floral earrings, hair and visible wrapped neckline from <Picture 4> throughout; <Picture 5> provides same-person side/back guidance only. Keep the narrow lower-garment silhouette from composition guidance. Neither identity image supplies scenery, sheet background, panels, labels or additional women.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the existing stair position and world-based sun and wind, interpreted through the actual side-rear view.
<Picture 1> ([Shot 1] stopped medium opening): partially_preserved - retain the defined framing, shoulder line and single-turn pose progression; facial and upper-costume identity follows the canonical <Subject 1>, with no added smile or lip-corner lift.
<Picture 2> ([Shot 1] pre-turn close composition): partially_preserved - retain the defined framing, shoulder line and single-turn pose progression; facial and upper-costume identity follows the canonical <Subject 1>, with no added smile or lip-corner lift.
<Picture 3> ([Shot 1] backward-gaze composition): partially_preserved - retain the defined framing, shoulder line and single-turn pose progression; facial and upper-costume identity follows the canonical <Subject 1>, with no added smile or lip-corner lift.

detailed_description:
Live-action skin shows natural pores and subtle colour variation. Preserve normal slender adult proportions and weighted vertical robe folds, blue-grey cloud shadows, neutral-grey stone, warm side-edge light only, and restrained silk/metal highlights. Avoid an overall red-orange wash, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open directly in the side-rear medium composition of <Picture 1>, following a clear editorial cut from the preceding far-wide ending. <Subject 1> is already stopped on the same stair lane; there is no camera flight from the remote view into this portrait. Her body still faces uphill. Show the canonical <Subject 1> from <Picture 4> immediately: the layered floral crown with its long bead-drop chains, shorter floral earrings, high dark updo and rear hair, and original wrapped neckline with exposed upper arms. <Picture 5> assists this side-rear view of the same woman. Keep her supporting stance settled.
From 00:00.000 to 00:00.400, let a small natural sleeve-fold adjustment and subtle visible breath give the stopped body life. Her fingers move one fold by only a few centimetres, then relax. She keeps the left-facing profile without stepping or starting the turn. This brief action leads into the camera approach.
From 00:00.400 to 00:01.300, push smoothly along the existing side-rear view toward the pre-turn close composition of <Picture 2>. Keep the same left-facing profile through this approach; the backward turn has not begun. Tighten through distance without changing body or face proportions. Allow only a small framing adjustment around the existing shoulder; do not orbit across the action axis.
From 00:01.300 to 00:02.700, she turns her head backward once toward the viewer. The neck initiates the movement and the near shoulder follows only a little, while the torso stays oriented along the uphill route. Her eyes arrive after the head begins turning. Keep the jaw and neck connected naturally, with the same facial proportions throughout.
The earrings lag briefly behind this one turn, make a small diminishing sway and begin to settle. Hair strands respond to the established breeze without repeatedly covering the eyes. World light and wind stay fixed, with screen directions following the actual side-rear angle.
From 00:02.700 to 00:03.600, decelerate the camera into the close composition of <Picture 3> as her gaze settles toward the lens. Let the eyes carry the restrained, magnetic quality of the moment. Keep her original relaxed lip shape, with no added smile, lip-corner lift, wink, blown kiss or extra hand performance.
From 00:03.600 to 00:04.000, retain the settled gaze for only this short final interval. Keep subtle natural facial life rather than a rigid held photograph. End after the one completed turn, with a quiet close portrait and complete silence.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, wind, ambient sound, or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman identified by <Picture 3>, with its original facial shape, natural skin, dark updo, gold-and-white flower ornament with small red beads, long pearl-chain earrings and ivory-and-gold embroidered neckline. Her lower gown keeps the same slender narrow silhouette visible in <Picture 1> and <Picture 2>; the reflected hem belongs to this same person.
<Subject 2> is the same hard grey-stone platform, ancient pine on the established left side, fixed circular architectural frame, distant giant moon and cloud landscape in <Picture 1> and <Picture 2>. The stone carries a shallow wet sheen, not a deep lake.
<Picture 1> is the low-camera composition of real gown fabric above its attached ground reflection at the opening of [Shot 1].
<Picture 2> is the broad composition approached by the diagonal retreat, small lateral glide, upward reveal and wider field of view in [Shot 1].

summary:
[reference generation] Generate one five-second low oblique reveal from the real hem and wet-stone reflection of <Picture 1> toward the complete ring, moon and pine composition of <Picture 2>. <Subject 1> stays planted and transfers weight slightly. The camera glides diagonally backward, lifts and widens the view with acceleration and deceleration; <Picture 3> supplies identity only.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the same original face, skin, hair, head ornament, earrings and neckline from <Picture 3>, with the narrow adult silhouette. The ground reflection remains optically attached to her, not a second person.
<Subject 2> (appears in [Shot 1]): fully_preserved - preserve platform solidity and the world positions of pine, ring and moon throughout the camera reveal.
<Picture 1> ([Shot 1] hem-and-reflection opening): fully_preserved - retain the real fabric above the stone and its corresponding reflection below it, without rotating the picture upside down.
<Picture 2> ([Shot 1] wide reveal): fully_preserved - retain the fixed landscape arrangement, circular frame and broad world scale as more of the existing space becomes visible.

detailed_description:
Live-action skin shows natural pores, fine lines and subtle colour variation. Preserve normal slender adult proportions, a narrow weighted robe silhouette, blue-grey cloud shadows and neutral-grey stone, with warm side-edge light and restrained silk/metal highlights. Avoid red oily skin, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open close above the wet stone in the low composition of <Picture 1>. Real ivory-and-gold gown fabric occupies the upper near frame, and its corresponding inverted reflection lies below on hard grey slabs. Keep the camera upright and above the floor. It neither rotates the image nor travels underwater, and the reflection never becomes another person.
From 00:00.000 to 00:00.700, <Subject 1> transfers her weight slightly within the same planted stance. Her soles stay on the existing slab points; the hem's contact area does not travel across the platform. A small shift in cloth tension alters one real fold and its attached reflection. Show physical fabric response without a new walking step, enlarged body or inflated skirt.
From 00:00.700 to 00:01.500, make a low diagonal retreat with a small lateral glide. Near stone grain and reflected light spots pass quickly along the lower edge, creating strong foreground parallax before the camera rises. Brief physical streaking may affect those nearest moving details, but keep the real woman and distant world readable; do not smear the image into painterly blur. The wet sheen remains shallow and continuous, without waves or splashes.
From 00:01.500 to 00:03.400, accelerate the diagonal withdrawal and crane upward, gradually widening the field of view to reveal the full world planned by <Picture 2>. This is an oblique ground-level reveal, not another purely axial pullback. Keep <Subject 1> at her established position and preserve a continuous change of perspective through stone seams, cloth and pine needles.
Her sole identity comes from <Picture 3>: the original face and skin, gold-white flowers with small red beads, long pearl-chain earrings and ivory-and-gold neckline. The additional identity picture never becomes a half-body insert or another angle. Her head stays gently low and her lips remain unsmiling.
As the camera rises, disclose the ancient pine on its established left side, the fixed circular architectural frame and the distant giant moon. The ring is already complete, not a door opening. Its apparent coverage and the moon's coverage change with field of view and cropping, never physical expansion. Near stone and pine layers move faster in perspective than the far moon; no landmark wanders independently.
From 00:03.400 to 00:04.400, decelerate the diagonal movement and rise into the broad composition. Keep the platform horizontal, ring orientation continuous and gown silhouette slender.
From 00:04.400 to 00:05.000, settle briefly with subtle natural cloth life after the weight transfer. There is one real woman and her optically attached reflection. End in a continuous silent shot, with a short usable wide landing for editing.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, water sounds, wind, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman identified by <Picture 3>, retaining its original face, natural skin, dark updo, gold-white flower ornament with small red beads, long pearl-chain earrings and ivory-and-gold neckline. Her normal slender proportions and narrow gown silhouette follow <Picture 2>; her already-held right-hand-over-left-sleeve pose is lightly adjusted once toward <Picture 3>.
<Subject 2> is the same wet hard-stone platform, fixed circular frame, giant moon, ancient pine and cloud landscape in all three references.
<Picture 1> is the opening broad composition for the lateral slide and shallow forward arc in [Shot 1].
<Picture 2> is the full-body composition passed through as the camera decelerates and the woman completes her small steps.
<Picture 3> is the final low-gaze half-body composition and single sleeve-gathering action reference, as well as the character-identity source.

summary:
[reference generation] Generate one five-second continuous lateral slide and shallow forward arc through the world of <Picture 1>, the full-body passage of <Picture 2> and the low-gaze half-body composition of <Picture 3>. <Subject 1> takes one or two small real steps, stops and smooths the already-held left sleeve once with her right hand. Foreground pine tips skim the edge outside her face, with no axis crossing.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - keep the original facial and costume identity from <Picture 3>, the slender narrow body silhouette and the calm low gaze without smiling.
<Subject 2> (appears in [Shot 1]): fully_preserved - keep the same platform and world positions of ring, moon and pine through the approach.
<Picture 1> ([Shot 1] broad opening): fully_preserved - retain its initial world layout and character placement.
<Picture 2> ([Shot 1] full-body passage): fully_preserved - retain normal adult proportions and narrow clothing while real steps occur; do not substitute a new static figure.
<Picture 3> ([Shot 1] low-gaze half-body ending): fully_preserved - retain the reference's original face, jewellery, neckline and right-over-left sleeve relationship, preserved through one small smoothing motion along the already-held sleeve, without releasing and grabbing again.

detailed_description:
Live-action skin shows natural pores, fine lines and subtle colour variation. Preserve normal slender adult proportions, a narrow weighted robe silhouette, blue-grey cloud shadows and neutral-grey stone, with warm side-edge light and restrained silk/metal highlights. Avoid red oily skin, waxy smoothing and painterly/cartoon shading.
[Shot 1] Open in the broad world of <Picture 1>. <Subject 1> is already the same single adult woman, with the original identity and clothing of <Picture 3>. Her gaze is low and her right hand already rests across the left sleeve near the waist. Do not lower her hands to restart the pose. The pine remains upper left, the moon behind toward frame right and the circular frame fixed in place.
From 00:00.000 to 00:00.500, start a restrained lateral slide that leads into a shallow forward arc toward her left-front. From 00:00.500 to 00:02.100, accelerate this approach while limiting the viewpoint change to roughly three to five degrees. Remain on the same side of the action axis, rather than making a half-circle around her. Appropriate gradual narrowing of the field of view brings the full-body composition nearer.
Use the left-upper pine needles and branch tips as close foreground layers. They skim the lens edge briefly, creating partial local occlusion and strong near/far parallax. Keep them outside her face and eyes; they neither touch her head nor make a full-frame black wipe. As these layers separate, her figure becomes clearer behind them. The pine itself does not move to manufacture the effect.
While the framing approaches <Picture 2>, she takes one or two small steps toward the camera on the hard slabs. Each shoe plants before weight transfers. These real steps move her foot contacts, hem and their surface reflections; do not simultaneously lock the feet to the starting points. The short advance leaves the scene layout intact. Keep her slender narrow gown moving around real legs without widening into a bulky skirt.
From 00:02.100 to 00:03.300, reduce the arc and approach speed as she finishes the last step and settles. The moon remains distant and the ring edge gradually leaves the crop; apparent image size follows the changing view and focal framing, never a growing celestial body or opening ring. Maintain the same original face, neckline and jewellery.
From 00:03.300 to 00:04.500, her right fingers glide once along the already-held left sleeve, gently smoothing one fold and returning to the resting relationship of <Picture 3>. She does not drop the arm, release and grab again, bow or repeat the adjustment. Her shoulders remain quiet and her eyes stay down.
From 00:04.500 to 00:05.000, decelerate into the low-gaze half-body framing. Let face detail and natural skin texture read clearly after the brief edge occlusion. Keep the completed sleeve pose and stopped feet for the next unit, with unmoving lip corners and one continuous silent shot.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, water sounds, wind, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman identified by <Picture 1>, with its original face, natural skin, dark updo, gold-white flower ornament with small red beads, long pearl-chain earrings and ivory-and-gold neckline. Her lips retain the same naturally relaxed shape and corner height during the action.
<Subject 2> is the same platform, ancient pine, fixed circular frame, giant moon and cloud world visible around her in <Picture 1> and <Picture 2>.
<Picture 1> is the opening low-gaze half-body pose, with planted feet and the right hand already holding the left sleeve after the preceding unit.
<Picture 2> is the closer composition and raised eye-direction guidance reached by a very small shoulder-side arc, modest portrait-field narrowing and one small head lift; it does not override identity or introduce a smile.

summary:
[reference generation] Generate one four-second portrait with a very small sideways shoulder arc from <Picture 1> toward the close scale of <Picture 2>. The camera settles early; <Subject 1> raises her head slowly and lets the eyes reach the lens afterward. Preserve cool melancholy magnetism, the same stopped sleeve pose and unchanged lips, without another straight face push.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the original facial identity, natural skin, ornament, earrings, neckline and relaxed lip shape from <Picture 1>, with the completed sleeve pose carried forward.
<Subject 2> (appears in [Shot 1]): fully_preserved - keep the same environment and fixed moon, ring and pine positions behind the slight side arc.
<Picture 1> ([Shot 1] settled low-gaze opening): fully_preserved - inherit the already-stopped body and completed right-over-left sleeve pose without resetting or repeating the previous actions.
<Picture 2> ([Shot 1] closer raised-gaze composition): partially_preserved - retain close framing and eye direction while using <Picture 1> for identity and unchanged lips; exclude smile-like lip-corner changes or a different facial template.

detailed_description:
Live-action skin shows natural pores, fine lines and subtle colour variation. Preserve normal slender adult proportions, a narrow weighted robe silhouette, blue-grey cloud shadows and neutral-grey stone, with warm side-edge light and restrained silk/metal highlights. Avoid red oily skin, waxy smoothing and painterly/cartoon shading.
[Shot 1] Begin in the low-gaze half-body pose of <Picture 1>. <Subject 1> has completed the previous steps and small sleeve adjustment: both feet are stopped, the right hand retains the left sleeve and her chin is gently lowered. Keep her original face and natural skin, gold-white flowers with small red beads, long pearl-chain earrings, dark updo and ivory-and-gold neckline unchanged.
From 00:00.000 to 00:01.100, make only a very small sideways arc around the near shoulder at nearly the same distance. A modest narrowing of the portrait field of view may complete the crop, but stop at the close composition of <Picture 2>; the crown already approaches the frame edge. Do not add a fast straight push into her face, continue toward the eye, cross the axis or rotate her shoulders to help the camera. The moon, ring and pine remain fixed behind the changing crop.
Her half-body face starts clear. As the camera settles, gently prioritise the eye plane without conspicuously defocusing the face. Keep the eyes readable through the movement and the final close view, with no dramatic rack-focus blur or apparent reshaping of the crown, jaw or cheeks. Camera movement finishes early so her own action carries the emotional change.
From 00:01.100 to 00:02.200, she slowly raises her head a little and straightens it only enough to address the lens naturally. The neck initiates the action; the jaw follows with its original proportions. Her shoulders, hands and supporting stance remain in place. The earrings lag once behind this small movement and settle with a diminishing sway. She does not walk again or gather the sleeve again.
From 00:01.800 to 00:03.400, her eyes rise to the camera after the head has begun lifting. Keep that slight delay purposeful and human. Her attention is direct but cool and emotionally distant, with a trace of melancholy and restrained magnetism. Do not substitute a cheerful invitation, exaggerated squint or theatrical sorrow. Her lips keep the naturally relaxed shape of <Picture 1>, including any original slight separation; neither corner rises and there is no speech-like movement.
From 00:03.400 to 00:04.000, retain the completed gaze for this brief final interval. The camera is steady at the selected close scale, without cropping farther into the crown or eyes. Keep the eyes sharp and quietly alive, with natural skin detail rather than a frozen blank portrait. No smile, wink, kiss gesture, crying or extra hand performance is added. Finish with one same adult woman in one continuous silent shot.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric sounds, water sounds, wind, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman whose original face, natural skin, red-and-cream floral crown and hanging gold/pearl ornaments come from <Picture 3>. Her deep-violet fine-silver-embroidered robe, slender adult proportions, narrow skirt and dry-bank foot placement follow <Picture 2>. The small distant face in <Picture 1> is not a different identity.
<Subject 2> is the same white deer in <Picture 1> and <Picture 2>, with pale fur and its original crystalline antlers, drinking at the left shallow water edge.
<Subject 3> is the same cloud-palace shore world with right-side light; different camera setups may place the palace differently within the frame without moving it physically.
<Picture 1> guides [Shot 1] broad space, scale and woman-right/deer-left layout only.
<Picture 2> guides [Shot 2] costume, full-body proportions, dry-bank foot position and deer relationship at the new camera setup after an explicit drinking-action match cut.

summary:
[reference generation] Generate a five-second two-shot entrance: [Shot 1] approaches diagonally with a modest descent from <Picture 1>; at 00:02.400, an explicit action-matched cut continues the same deer sip into [Shot 2], the full-body composition of <Picture 2>. Palace frame-position change belongs to the camera-setup change, not background morphing. <Picture 3> supplies the original woman identity throughout.

retention_analysis:
<Subject 1> (appears in [Shot 1], [Shot 2]): fully_preserved - retain face and crown from <Picture 3>, with the same purple-silver robe, adult proportions, hand pose and dry-bank stance from <Picture 2>.
<Subject 2> (appears in [Shot 1], [Shot 2]): fully_preserved - retain the same deer, crystal antlers and single continuous small drinking action through the cut.
<Subject 3> (appears in [Shot 1], [Shot 2]): fully_preserved - retain the physical world and right-side light; different palace framing is explained by the explicit camera-setup edit.
<Picture 1> ([Shot 1] broad composition): partially_preserved - retain wide space, scale and subject directions only; use the canonical face and do not morph this setup into <Picture 2>.
<Picture 2> ([Shot 2] full-body composition): fully_preserved - retain its defined costume, normal proportions, planted dry-bank feet and woman-right/deer-left layout.

detailed_description:
Live-action fantasy imagery with natural pores, fine skin texture and normal slender adult proportions. Keep weighted narrow purple-silver fabrics, pale deer fur, neutral-grey stone and blue-grey shadows under thin warm light from the right; avoid oily red skin, plastic smoothing and painterly/cartoon shading.
[Shot 1] Open in the broad shore composition of <Picture 1>. <Subject 1> stands on the established dry right bank, with <Subject 2>, the same white deer, drinking on the left. Her original face and floral crown follow <Picture 3>, never a guessed face enlarged from the distant image. Her deep-violet silver-embroidered narrow robe follows <Picture 2>. The cloud palace, water and stone retain the broad setup's spatial layout.
From 00:00.000 to 00:00.700, establish real small motion. Her quiet breath slightly changes fabric tension while the waist-level hand pose and planted feet stay unchanged. The deer remains head-down, with its original antlers attached to that same head. Reflections belong to the existing bodies, not extra beings.
From 00:00.700 to 00:02.000, move the camera diagonally toward the pair and descend only modestly. Keep the horizon level. Nearby rock edges and antlers show more parallax than the distant palace. Do not force the wide setup's palace coordinates to match the full-body view, and do not use a high aerial plunge or flat image enlargement.
From 00:02.000 to 00:02.400, the deer makes one small muzzle contact or sip at the water surface. Its jaw and neck respond a little, with a local ring beneath the nose. Use this action phase as the edit point, without lifting the head high or inventing antler branches.
[Shot 2] At 00:02.400, explicitly action-match cut to the full-body composition of <Picture 2>, continuing that SAME muzzle movement and ripple phase. This is a different camera setup in the same world. The palace now appears farther toward frame left because the viewpoint changes; it never slides physically and the bank does not deform.
From 00:02.400 to 00:03.500, settle a very short residual camera movement. The woman is already complete on her original dry-bank foot points, with unchanged hands and the original face from <Picture 3> clear at this scale. Keep normal adult height, a weighted narrow skirt and the same purple-silver embroidery. She neither enters the pool nor becomes a shortened body.
From 00:03.500 to 00:04.400, finish the same gentle drinking action without restarting the sip. The deer remains at the left shallow edge as the small ripple softens; her visible breath and fine cloth response keep the stopped body alive.
From 00:04.400 to 00:05.000, retain a brief grounded full-body landing. Pale fur, skin and silver threads remain textured under thin warm right-side light and blue-grey shadows. Both shots are completely silent.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, hoof or foot sounds, fabric sounds, drinking sounds, water sounds, wind, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman whose original facial identity, natural skin, floral crown and hanging ornaments come from <Picture 3>. Her deep-violet silver-embroidered robe, slender proportions, dry-bank stance and existing hand relationship follow <Picture 1>. Her close gaze is directed down-left toward the deer as framed in <Picture 2>.
<Subject 2> is the original white deer and its attached crystalline antlers at the lower-left foreground in <Picture 1>, still drinking at the same water edge.
<Subject 3> is the same shore, water, cloud-palace background and right-side light in <Picture 1> and <Picture 2>.
<Picture 1> is the full-body composition for [Shot 1], with real deer antlers at the lower-left foreground and the woman at the right.
<Picture 2> is the already-existing close portrait selected by a motion-matched cut into [Shot 2], not a person generated out of an antler image.

summary:
[reference generation] Generate a five-second antler-led shot with one explicit internal edit. In [Shot 1], attention briefly reaches <Subject 2>'s real lower-left-foreground crystal antlers, then a fast level pan whips toward the woman on the right. At 00:01.200, a motion-matched cut enters [Shot 2] in the close composition of <Picture 2>; <Subject 1> is already looking down-left toward the deer, and the camera immediately settles.

retention_analysis:
<Subject 1> (appears in [Shot 1], [Shot 2]): fully_preserved - retain the original face and crown from <Picture 3>, the same purple-silver robe and unchanged dry-bank stance; the edit changes framing rather than materialising a new woman.
<Subject 2> (appears in [Shot 1], outside the close crop in [Shot 2]): fully_preserved - retain the same one deer, existing antlers and drinking position; antlers stay on the left and never become a new foreground object across her face.
<Subject 3> (appears in [Shot 1], [Shot 2]): fully_preserved - retain the same location, shore and right-side light across the matched cut.
<Picture 1> ([Shot 1] full-body starting view): fully_preserved - retain the grounded body, existing antlers and woman-right/deer-left layout.
<Picture 2> ([Shot 2] close portrait): partially_preserved - retain close framing and down-left deer-directed gaze while the canonical face stays from <Picture 3> and any visible cloth stays deep purple-silver from <Picture 1>.

detailed_description:
Live-action fantasy imagery with natural pores, fine skin texture and normal slender adult proportions. Keep weighted narrow purple-silver fabrics, pale deer fur, neutral-grey stone and blue-grey shadows under thin warm light from the right; avoid oily red skin, plastic smoothing and painterly/cartoon shading.
[Shot 1] Begin at the full-body woman-and-deer composition of <Picture 1>. <Subject 1> is the same adult woman on the right dry bank, with original face and floral crown from <Picture 3>, wearing the deep-violet silver-embroidered robe. Her feet and waist-level hands retain their existing positions. <Subject 2> drinks on the left; only its original attached crystalline antlers create the lower-left-foreground layer.
From 00:00.000 to 00:00.600, briefly rack focus onto the real lower-left crystalline antler facets, while keeping the complete woman recognisable behind them. The deer makes a tiny muzzle or jaw movement and a local water ring. Do not invent antlers near her face or a refracted person inside a crystal.
From 00:00.600 to 00:01.200, whip the camera horizontally toward frame right, in the direction of the woman. Keep the pan level, without a full-circle orbit or passage through the antlers. The original lower-left antler layer slips toward the left edge and out of view as the pan accelerates. Genuine short motion blur follows this fast camera movement; it does not melt the scene or transform deer anatomy into skin.
[Shot 2] At 00:01.200, make an explicit motion-matched edit to the close composition of <Picture 2>, continuing the same horizontal movement for an instant. <Subject 1> is already fully present, already looking down-left toward the deer. This is a new camera setup at the same bank position, not an unbroken flight from an antler into a human or a body appearing from its reflection.
From 00:01.200 to 00:01.700, brake the residual camera motion quickly into the selected close framing. Immediately lock focus on her face and eyes as the brief motion-blur interval ends. Preserve the original facial proportions, natural skin detail, crown and individual hanging chains from <Picture 3>; avoid an oily red face or a blurred painterly surface.
From 00:01.700 to 00:04.400, let the calm deer-directed gaze remain alive. A slight natural head inclination and quiet visible breath give the portrait human weight. The earrings and crown chains show one tiny delayed sway and settle. Hands and feet remain in their original relationship outside the crop; she does not walk, bow or perform another large face-directed push.
From 00:04.400 to 00:05.000, retain a short settled landing for the next gaze change. If garment edges are visible, they stay the same purple-silver cloth; do not widen the portrait merely to show a robe. The deer remains outside this close crop at its original left position. Both shots remain completely silent.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, hoof or foot sounds, fabric sounds, drinking sounds, water sounds, wind, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman whose original face, natural skin, floral crown, nose and lip shapes, and hanging ornaments follow <Picture 2>. Her deep-violet robe with fine silver embroidery follows <Picture 3> only where cloth is naturally visible. The selected close crop does not expand to force the robe into view.
<Subject 2> is the existing white deer she was watching in the down-left gaze of <Picture 1>, remaining outside the portrait to her left; no new animal or antler enters this close crop.
<Subject 3> is the same cloud-palace shore world with main light from the right, seen behind the close compositions of <Picture 1> and <Picture 2>.
<Picture 1> is the close down-left deer-directed gaze at the opening of [Shot 1].
<Picture 2> is the original face and frontal close composition reached by an eye-first attention change followed by a small head adjustment.

summary:
[reference generation] Generate one four-second mostly steady close portrait from <Picture 1> toward <Picture 2>. <Subject 1>'s eyes leave the down-left deer and find the lens first; her head then follows slightly toward the front. Keep her body and hands at the same bank position, with natural restrained magnetism rather than another head-lift-first performance. <Picture 3> supplies costume continuity only.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the original face, skin, crown and relaxed lip shape from <Picture 2>, with purple-silver costume continuity from <Picture 3> outside or at the natural crop edge.
<Subject 2> (outside the crop during [Shot 1]): fully_preserved - retain the same left-side drinking deer as the existing gaze target, without adding antlers across the portrait.
<Subject 3> (appears in [Shot 1]): fully_preserved - keep the same location and right-side light while attention changes.
<Picture 1> ([Shot 1] initial close gaze): partially_preserved - retain down-left eye direction and close framing; keep facial identity and any visible garment aligned with the canonical <Subject 1>.
<Picture 2> ([Shot 1] frontal close ending): fully_preserved - retain original face, floral crown and close scale with a natural eye-first transition, without requiring a smile.

detailed_description:
Live-action fantasy imagery with natural pores, fine skin texture and normal slender adult proportions. Keep weighted narrow purple-silver fabrics, pale deer fur, neutral-grey stone and blue-grey shadows under thin warm light from the right; avoid oily red skin, plastic smoothing and painterly/cartoon shading.
[Shot 1] Open in the close down-left gaze composition of <Picture 1>. <Subject 1> is already stopped on the same dry bank, looking toward the white deer outside the left edge. Her body, shoulders and waist-level hands keep the stance established in the full-body reference. Use <Picture 2> for the original face, natural skin, floral crown and hanging ornaments; use <Picture 3> only for purple-silver costume continuity if an edge of fabric is naturally visible.
From 00:00.000 to 00:00.500, let the existing deer-directed attention register briefly. The camera stays at close portrait distance and nearly steady, with no new fast push, pullback or large side orbit. Her lips keep their naturally relaxed source shape, including any slight separation. The head is already at the needed portrait height; she does not repeat another character's chin-lift-first action.
From 00:00.500 to 00:01.500, her eyes move from the down-left deer toward the lens before the head starts to follow. Let the change begin as a small, purposeful shift in attention rather than a sudden wide-eyed snap. The pupils and eyelids retain the original eye shape, and the brows remain relaxed. Keep the eye plane clear without a dramatic rack focus.
From 00:01.500 to 00:02.800, the head follows that new attention by only a few degrees toward the frontal composition of <Picture 2>. The neck and jaw remain natural, with no face slimming, elongated neck or different identity. Her shoulders do not rotate into another stance. The earrings and crown chains lag slightly behind this single small turn, then settle once.
Keep the camera mostly anchored during the eye-first and head-follow action. A tiny compositional correction may keep the eyes comfortably placed, but the motion does not orbit beyond the original close-view range or reframe to a full body. The deer's existing left-hand location and the main light from the right remain unchanged behind this attention shift.
From 00:02.800 to 00:03.400, let the gaze settle toward the viewer. The expression has quiet curiosity and restrained, natural magnetism, carried by living eyes and a softly relaxed face. Do not demand a broad smile or raised lip corners, and do not make the portrait a rigid blank stare. There is no exaggerated wink, kiss gesture or speech-like mouth action.
From 00:03.400 to 00:04.000, keep this short frontal-gaze landing with natural skin detail and subtle facial life. Preserve the selected crop: if no robe is visible, do not add purple fabric over bare shoulders or reveal a new outfit. Finish with the same single woman, grounded body and completely silent continuous shot.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, hoof or foot sounds, fabric sounds, drinking sounds, water sounds, wind, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman appearing only after the cut in [Shot 2], whose original face, crown, black hair and long gold earrings follow <Picture 3>, with the same white fur collar, embroidered costume and waist-level hands in <Picture 2>. The preceding empty shot contains no person.
<Subject 2> is the cloud-cascade and palace world in <Picture 1>. Existing white cloud volumes subtly flow DOWN, with lit ridges and darker folds; the camera's upward movement is separate. Clouds are not water or glowing liquid, and the palace roofs, eaves and columns remain rigid in the world.
<Picture 1> is the unoccupied establishing composition for [Shot 1], using its existing LEFT cloud cascade as the upward-camera path and texture-matching subject.
<Picture 2> is the complete frontal half-body composition selected by an explicit cloud-texture cut into [Shot 2], retaining left cloud cascade and palace right-behind.

summary:
[reference generation] Generate a five-second two-shot introduction. [Shot 1] follows the existing LEFT cloud cascade with an upward tilt and modest rising glide: slow start, faster rise against subtly descending clouds, then deceleration onto the cloud texture. At 00:03.000, explicitly cloud-texture match cut to [Shot 2], the already-complete half-body woman of <Picture 2>; carry only a short upward camera direction before quickly settling. <Picture 3> supplies identity only.

retention_analysis:
<Subject 1> (appears only in [Shot 2]): fully_preserved - retain the original face, crown, hair and earrings from <Picture 3>, with white fur, costume and original waist-level hands from <Picture 2>; she is absent before the cut and never grows from clouds.
<Subject 2> (appears in [Shot 1] and the background of [Shot 2]): fully_preserved - retain subtly descending volumetric clouds, rigid architecture and photographic depth; upward camera motion does not stretch or independently translate the palace.
<Picture 1> ([Shot 1] empty scene and left cloud texture): fully_preserved - retain the absence of people and existing left cascade; distant cloud texture is composed optically, without invented near-lens fog or a foreground wipe.
<Picture 2> ([Shot 2] half-body landing): fully_preserved - retain one complete woman, original proportions, fur and hand relationship, with cloud left and palace right-behind; the brief continued upward direction is camera motion after a real edit.

detailed_description:
Live-action photographic imagery with natural pores, fine skin texture and normal adult proportions. Keep neutral skin, blue-grey surroundings, soft side fill, tactile white fur and restrained silk/metal highlights; avoid oily red colour, waxy plastic smoothing and painterly/cartoon shading.
[Shot 1] Open in the unoccupied cloud-palace composition of <Picture 1>. The existing LEFT cloud cascade is the main vertical path, with white rounded volumes, lit crests and cool darker folds. Hard roofs and columns provide scale. There is no woman, robe, human silhouette or forming body in the scene, including the lower foreground.
From 00:00.000 to 00:00.500, begin slowly with a brief broad reading of the cascade. Cloud layers subtly flow DOWN; they remain cloud volume, not falling water, splashes or glowing liquid. The palace is fixed in world space rather than carried by the clouds.
From 00:00.500 to 00:01.200, tilt gently UP along the left cascade and add a light rising glide. Let lower folds give way to higher cloud layers and palace eaves through a coherent viewpoint change. Keep roof shapes and column lengths intact.
From 00:01.200 to 00:02.300, accelerate the upward tilt and modest rise. The camera travels UP while the cloud material continues subtly DOWN, producing stronger opposing screen motion and a clear sense of height. Do not exaggerate this into a vast flight or stretch the cloud architecture. Palace movement within the frame comes only from the changing view, not independent translation.
From 00:02.300 to 00:03.000, decelerate into the existing left cloud texture as the dominant composition, using suitable field-of-view narrowing if needed. Preserve its white crests and grey recesses. This is optical framing of established distant texture, not new close fog brushing the lens, a dive through clouds or a fabricated foreground wipe.
[Shot 2] At 00:03.000, make an explicit cloud-texture match cut to the frontal half-body composition of <Picture 2>. The camera setup changes; the woman is already fully complete in the first image, with the original face and gold crown from <Picture 3>, black hair, long gold earrings and natural white fur. She does not emerge, fade in or grow from vapour, and this is not continuous travel from remote clouds to her body.
From 00:03.000 to 00:03.300, carry a short restrained upward camera direction into this new setup, then brake quickly. Keep cloud cascade on the left and rigid palace right-behind, without stretching or shifting the buildings.
From 00:03.300 to 00:04.300, she makes only a slight weight adjustment and quiet visible breath. Fur and fabric respond gently; both hands retain the original waist-level relationship. Do not add a step, large head turn or fur-arranging gesture in this unit.
From 00:04.300 to 00:05.000, keep the settled half-body landing, neutral skin and softly clear gaze. Prepare her following single-step approach with no extra performance. Both shots remain completely silent.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric or fur sounds, wind, cloud sounds, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.

subject_definitions:
<Subject 1> is ONE adult woman with the original face, crown, black hair and long gold earrings of <Picture 2>, retaining the white fur collar, embroidered costume and original waist-level hand relationship from <Picture 1>. She takes only one natural small step toward the camera.
<Subject 2> is the same cloud-cascade and palace background in <Picture 1> and <Picture 2>, with descending volumetric cloud layers to the left and rigid palace structure behind on the right.
<Picture 1> is the frontal half-body starting composition and unchanged hand relationship for [Shot 1].
<Picture 2> is the selected frontal close composition, reached mainly by the woman's actual approach and only a modest portrait-field adjustment; the camera stops at this crop rather than pushing beyond the crown edge.

summary:
[reference generation] Generate one four-second actor-led close approach from <Picture 1> toward <Picture 2>. <Subject 1> advances one small natural step, with visible body and garment inertia; a short responsive camera adjustment and modest field narrowing complete the close framing. She stops, meets the lens with soft restrained magnetism and permits only a minute natural relaxation at the lip corners, without a large smile.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the same original face, crown, hair, earrings and white fur, with original waist-level hands and one genuine body advance rather than a scaled still portrait.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain cloud cascade on the left, palace right-behind and the same light during the change in portrait scale.
<Picture 1> ([Shot 1] half-body opening): fully_preserved - retain original adult proportions, fur placement, costume and hand relationship at the start of the one step.
<Picture 2> ([Shot 1] close landing): fully_preserved - retain facial identity and selected close scale, without continuing into an eye macro or forcing a wider frame to show off-screen feet.

detailed_description:
Live-action photographic imagery with natural pores, fine skin texture and normal adult proportions. Keep neutral skin, blue-grey surroundings, soft side fill, tactile white fur and restrained silk/metal highlights; avoid oily red colour, waxy plastic smoothing and painterly/cartoon shading.
[Shot 1] Begin at the frontal half-body scale of <Picture 1>. <Subject 1> is the same single adult woman with original face and crown from <Picture 2>, black hair, long gold earrings, white fur collar and the embroidered costume. Her hands remain in their loose original waist-level relationship. The background keeps the cloud cascade left and palace right-behind.
From 00:00.000 to 00:00.500, prepare one small step through a natural change in body support. Let the torso and shoulders respond softly while the hands stay composed. Her head stays mostly frontal; she does not combine the approach with a large turn, a bow or repeated fur handling. The lower legs are below the crop, so the visible action is carried by genuine weight transfer and upper-body travel.
From 00:00.500 to 00:02.000, she takes that ONE short step toward the camera. Her body actually advances and rises or settles slightly with normal balance, rather than becoming a larger static face through image scaling. The sleeve edges and fur show a small delayed response to the movement; loose hair and hanging earrings lag once and follow naturally. Keep the collar seated on the same shoulders and the face's proportions unchanged.
The camera records her approach with only a short responsive following adjustment to keep the portrait placed. Most tightening comes from her real movement. If needed, narrow the portrait field of view gently for the remaining crop, without another large forward drive, whip-pan, half-circle orbit or floor-level slide. The distance change stays physically readable through shoulders, clothing and background perspective.
From 00:02.000 to 00:03.100, she completes the step and comes to a settled stop. Let the shoulder and torso movement finish before adding expression. The camera eases into the close composition of <Picture 2>, stopping when the crown reaches the selected top-edge relationship. Do not continue closer into the eyes or invent extra headroom by changing the crown.
From 00:03.100 to 00:03.500, her gaze rests on the lens with quiet, gentle magnetism. The lip corners may soften once by a minute amount while keeping the naturally relaxed original lip shape. Do not reveal teeth, make a broad laugh or turn this into a wink, kiss gesture or speech-like mouth action.
From 00:03.500 to 00:04.000, retain the calm close landing. Fur, sleeves, earrings and fine hair finish their small inertial settling; her hands stay at the existing waist-level position outside or near the crop edge. Keep natural skin pores, neutral colour and softly clear eyes, not oily red gloss or plastic smoothing. The background cloud volume descends without becoming liquid, and the rigid palace does not drift. Finish one continuous entirely silent shot.

overall_soundscape:
Complete silence throughout: no dialogue, narration, singing, breathing sounds, footsteps, fabric or fur sounds, wind, cloud sounds, ambient sound or any sound effects.

non_diegetic_music:
None. No background music, score, beat or musical accent.