Đêm sinh nhật: chiếc bánh và một tin nhắn

Prompt gốc (nguyên văn)

[SPEC] 12 seconds, 9:16 vertical, 24fps, 1080p, realistic short drama, restrained emotion, no dialogue.

[CHARACTER LOCK] A clearly adult 26-year-old East Asian woman with distinctly East Asian features: a soft oval face, polished daytime makeup; a beige wool coat, cream sweater, black skirt, ankle boots and a light grey scarf. Keep the same face, outfit, hairstyle and body proportions for the entire clip.

[SETTING & LIGHT] A winter night in the stairwell of an older apartment block: concrete steps, green lower wall paint, a motion-sensor light, a security door covered in small flyers. She holds a white cream cake with lit candles, the flames lighting her face from below.

[SHOT LIST]
0-2.4s: she climbs the stairs carefully, one hand shielding the cake from drafts, steps light, an expectant smile on her lips.
2.4-4.8s: at a door she stops, takes a breath, knocks with her free hand, then leans her ear close to listen.
4.8-6.4s: no answer. She knocks again; the candles flicker in the draft and she turns her head away from the flames.
6.4-8.4s: she frees one hand to take out her phone and types single-handed, the screen light falling on her chin; the phone screen stays defocused and unreadable.
8.4-10s: the phone buzzes. She looks down. The smile is still on her face but her eyes stop, the smile fades little by little, brows tighten slightly, and she freezes in place.
10-12s: she slowly sets the cake down on the floor by the door, crouches and blows out the candles one by one. As the last goes out the frame darkens to the white motion-sensor light; she looks up at the door number, not crying.

[TEXTURE] Real handheld, slight noise in low light, the motion-sensor light timing out after a few seconds and being triggered again by footsteps; no lit-stage feel, no theatrical lighting.

[SOUND] Ambience only: stairwell echo, plastic bag, lighter, candle flames in a draft, phone vibration, distant TV. No music, no dialogue, no voice-over.

[NEGATIVE PROMPT] melodramatic crying, expression collapse, readable text on the phone screen, Western/Caucasian bone structure, non-Asian face, cake deformation, candle count changing, face swap, outfit change, corridor layout changing, extra people, music, subtitles, watermark.