subject_definitions:
<Subject 1> is the same adult woman from <Picture 1>: the exact referenced face, eyes, long dark hair arranged with the original flower-and-pearl ornaments, dangling earrings, cream patterned robe and restrained red embroidered borders. <Picture 1> is an identity and costume reference, not the scene background.
<Subject 2> is the same adult man from <Picture 2>: the exact referenced face, long dark hair and topknot with the original pin, dark finely patterned robe, subdued collar borders and pale inner collar. <Picture 2> is an identity and costume reference, not the scene background.
<Picture 3> is the composition, starting-pose, scene, lighting and color anchor for [Shot 1].

summary:
[reference generation] An 8-second single-shot close two-person ending anchored by <Picture 3>. <Subject 1> first continues looking upward, then turns toward <Subject 2>, establishes mutual eye contact, and only then speaks her final confession. Only the woman speaks.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement, ornaments and costume; allow only the specified performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement and costume; allow only the specified performance.
<Picture 3> ([Shot 1] composition and starting-pose anchor): fully_preserved - retain its scene layout, initial positions, framing, existing illumination and color relationship. The specified later movement develops from this anchor.

detailed_description:
The target uses the same visual treatment as the supplied scene still. Retain the source still's exact existing exposure, white balance, skin and fabric colors, saturation, contrast, shadow direction and background structure. Keep these stable over time; do not add warm lighting, orange grading, a filter, bloom, flare, haze or an exposure transition. Preserve realistic skin and fabric texture, the established woman-left/man-right geography and original close spacing. Use a locked camera, only subtle breathing and minute natural hair movement. No extra person, prop, embrace, subtitles or on-screen text.
[Shot 1] Begin in the close two-face framing of <Picture 3>. Both complete faces remain readable. <Subject 1> is on the left, still looking upward with the same head angle as before the cut. <Subject 2> is on the right and slightly behind, already looking at her face. Keep their torso positions and interpersonal spacing fixed. Use a natural young-adult female Mandarin speaking voice in a normal mid register: clear, gentle and rounded, with only a little natural breath. Keep the same underlying timbre, habitual pitch range, accent and articulation throughout the sequence. Emotion changes the breath, pauses and slight tremor, not the vocal identity. She is speaking to the man beside her, not narrating to an audience.
From approximately 00:00.300 to 00:03.100, <Subject 1> (S1) says, <d>[Chinese] 同样你也要记得。</d> Her tone is soft, serious and personal, gently asking him to remember. Keep her gaze UPWARD for this entire phrase. Deliver the words in one natural breath and let the last two syllables settle clearly. 
From approximately 00:03.100 to 00:04.300, she falls silent, lowers her upward gaze and slowly turns her head toward the man on her right. Her body stays still. He continues watching her and makes only the minute eye adjustment necessary to meet her gaze. Finish the turn and establish clear MUTUAL EYE CONTACT before the next line begins. This must be a continuous natural head movement, without a morph, jump, camera cut, embrace or added touch.
From approximately 00:04.400 to 00:06.400, while looking directly at his face and into his eyes, <Subject 1> (S1) says, <d>[Chinese] 我爱你。</d> Give the complete three-syllable phrase a full two-second speaking window in one connected breath. It is tender, definite and slightly choked, not a whisper that loses consonants. The final syllable must be fully voiced; do not cut it off, stretch each character separately or begin the line while she is still looking upward.
From approximately 00:06.400 through 00:08.000, hold the mutual gaze and allow a quiet, wordless breath after the completed confession. The man's mouth stays closed. Preserve the original face shapes throughout the turn, keep earrings attached naturally, and maintain stable illumination and source color through the final frame.

overall_soundscape:
A very faint, steady outdoor breeze beneath close, intelligible dialogue, with natural breathing and minimal fabric rustle. Maintain the same low ambience level across the sequence. No additional voices, vocal improvisations or conspicuous sound effects.

non_diegetic_music:
N/A
