Tang Dynasty Beauty Garden Video

FIRST FRAME (=Upload reference image, used only for initial pose, character moves naturally afterwards):

Tang Dynasty beauty with dark hair in a high bun, a red peony pinned next to a golden step-shake hairpiece, flower ornament on the forehead.

Wearing an off-white sheer gauze outer cape over a crimson and orange gold-traced tube top ruqun dress, hands naturally folded in front of the body.

Peony bushes nearby, Forbidden City towers in the background, warm golden sunset light, fine dust floating in the air.

Character features, makeup, and attire are consistent with the first frame; limbs remain naturally relaxed and not stiff throughout.

Core Requirement: Documentary-style live-action performance texture, removing all "AI flavor"—skin with real pores and fine fuzz, oily sheen, and visible rhythmic breathing; micro-expressions with imperfect realism (occasional eye wandering, corner of the mouth moving slightly before smiling, irregular blinking rhythm, occasional subtle movements like swallowing, pursing lips, or shifting weight).

STORYBOARD (Minimal dialogue, emphasis on realistic detailed performance):

SHOT 1 | 0-4s | Medium Close-up (Handheld, slight breathing shake)
She looks down at her folded hands, fingertips unconsciously fiddling with her belt, shifting weight from one foot to the other, shoulders relaxing; then drawn to the peonies beside her, naturally looking up and sideways, eyes leading the head turn.
No dialogue, only natural breathing and environmental sounds (wind, distant birds).

SHOT 2 | 4-7s | Face Close-up (Eye acting)
Camera pushes toward the face: irregular blinking, subtle focus shifts in the pupils, nostrils fluttering slightly with breath, lips pursing briefly before a real faint smile spreads, the smile starting from the eyes—not a formulaic standard smile.
[Barely a whisper, one line] "These flowers... are blooming so well."

SHOT 3 | 7-11s | Waist-up action
She reaches out to touch a peony petal, finger pads sinking realistically into the softness; wrist turns naturally, gauze sleeve sliding down, she subconsciously pulls it back with the other hand, body turning slightly, stray hairs blown by the wind sticking to her cheek. Frame is full of real, unintentional small movements. No dialogue.

SHOT 4 | 11-15s | Return to Medium Close-up (Closure, natural freeze)
She pulls back her hand, gaze returning forward, exhaling softly, eyes carrying a touch of realistic melancholy yet relief; lips move as if to say something but ultimately stop, ending with a smile that isn't perfect but very real, naturally freezing at that moment.
[Whisper, fading, like to herself] "...that's fine too."

AUDIO SETTINGS: Female voice, natural and lifelike, mostly breathy, speed varies with emotion, not overly articulated, with realistic breathing pauses; only two very short whispers throughout.

CAMERA: Handheld texture (subtle natural shake and breathing feel), multi-shot: MCU / Face Close-up / Half-body / back to MCU, natural transitions; realistic shallow depth of field, occasional subtle focus breathing, background palaces and peonies blurred.

ATMOSPHERE/LIGHTING: Natural golden hour light, rim lighting on hair, warm tones; floating dust, occasional falling petals, gauze and hair moving with a real breeze, overall look like film footage, fine grain, soft highlights.

CONSTRAINTS:
- 4K Ultra HD, 15 seconds, makeup/clothing consistent with the first frame.
- Core: Live documentary performance—real skin texture, irregular blinking, natural weight shifting, micro-expressions with "imperfect realism," strictly no stiffness or plastic faces.
- De-AI: Avoid excessive skin smoothing, perfect symmetry, or perfect smiles.
- Two lines strictly synced with mouth movements, breathy/lifelike, no broadcasting voice.
- Wind, falling flowers, sleeves, and hair follow real physics.
- Eyes must be expressive with real focus wandering; no glowing eyes or abnormal pupils.
- Ends naturally with residual movement, no black screen or transitions.
- No subtitles, watermarks, symbols, or text.

首帧 FIRST FRAME（=上传参考图，仅作起手姿态，之后人物自然活动）：

唐风美人，乌发高髻、金步摇旁簪一朵红牡丹，眉心花钿，

米白薄纱外披罩绛橙描金抹胸襦裙，双手在身前自然交叠；

身畔牡丹丛、身后紫禁城楼台、黄昏暖金光、空气微尘浮动。

人物五官妆容服饰与首帧一致，肢体全程自然松弛，不僵直。

核心要求：真人纪实表演质感，去掉一切「AI 味」——

皮肤有真实毛孔与细绒、油光与呼吸起伏，

微表情有不完美的真实感（眼神偶尔游移、嘴角先动一下再笑、

眨眼节奏不规律、偶有极轻的吞咽/抿唇/调整重心的小动作）。

分镜脚本（少台词·重真实细节表演）：

SHOT 1｜0-4秒｜中近景（手持镜头·极轻微呼吸晃动）

她低头看着交叠的双手，指尖无意识地轻捻衣带，

重心从一只脚换到另一只，肩膀松一下；

随后被身旁牡丹引去注意，自然抬眼侧望，眼神先到、头才转过去。

无台词，仅自然呼吸与环境声（风、远处鸟鸣）。

SHOT 2｜4-7秒｜面部特写（眼神戏）

镜头轻推到面部：眼睫不规律地眨动，瞳孔有细微对焦的游移，

鼻翼随呼吸极轻翕动，唇角先抿了一下、才漾开一个真实的浅笑，

笑意先从眼睛起来——不是程式化的标准微笑。

〔几乎气声·一句〕"这花……开得真好。"

SHOT 3｜7-11秒｜半身动态

她抬手轻触身旁牡丹花瓣，指腹真实地陷进花瓣的软，

手腕自然转动，纱袖滑下一截，她下意识用另一只手把袖子撩回去，

身子随之微微侧转，碎发被风吹乱、贴在脸颊，她不去理。

满画面都是真实的、不刻意的小动作。无台词。

SHOT 4｜11-15秒｜回到中近景（收束·自然定格）

她收回手，目光重新落回前方，轻轻吐了口气，

眼神里带着一点真实的怅然又释然，嘴唇微动想说什么、终究没说，

最后是一个没那么完美、却很真实的浅笑，自然停在那一刻。

〔气声·渐弱·像自语〕"……也好。"

声音设定：女声，自然生活化，气声偏多，语速随情绪不匀，

不字正腔圆、略带真实的呼吸停顿；全程仅两处极短轻语。

镜头 CAMERA：

手持质感（极轻微自然晃动与呼吸感），多分镜：

中近景／面部特写／半身／回中近景，切换自然不炫技；

真实浅景深、焦点偶有极细微的呼吸式重对焦，背景宫阙牡丹虚化。

氛围·光线：

黄昏真实自然光，侧逆光镶发丝金边，暖调；

空气微尘、花瓣偶尔飘落、纱披与碎发随真实微风动，

整体像胶片实拍，颗粒细腻、高光柔和不过曝。

CONSTRAINTS（约束）：

- 4K 超清，15秒，人物妆容服饰与首帧一致（换镜不换脸不换装）

- 核心：真人纪实表演——真实皮肤质感、不规律眨眼、自然重心转移、

微表情有「不完美的真实感」，严禁僵硬呆滞或塑料假脸

- 去 AI 味：避免过度磨皮、过度对称、过度顺滑的运动与完美微笑

- 两处台词与口型严格同步，气声生活化，不假唱、不字正腔圆

- 风动、落花、纱袖、碎发符合真实物理，自然不夸张

- 眼神有戏、有真实焦点游移；禁止眼神发光、瞳孔异常

- 结尾在自然的小动作余韵里定格，无黑屏无转场

- 无字幕、无水印、无任何标记符号、无文字