3D scene of a son feeding his elderly father with a…

SpringWangGPT Image 2.5

PROMPT

生成一张 9:16 竖屏、高品质三维动画电影质感的画面，作为一段温暖亲情短片的首帧。只生成一张完整画面，不要拼图、分镜板、字幕或播放器界面。

画面讲述同一对父子在两个年代里的照顾与被照顾：下方是现在，成年儿子正在喂年迈的父亲；上方是墙面上的往日回忆，年轻的父亲正在喂小时候的儿子。两组人物同时出现在同一个固定构图中。

场景是一间安静、温馨、有生活痕迹的老房间。暖褐色墙壁、深色木质护墙板，一张柔软的焦糖棕色高背扶手椅。傍晚的蜂蜜金色光线洒进房间，空气中只有少量细微浮尘。整体明亮柔和，人物面部和衣物细节清晰可见，阴影温暖通透，不要昏暗、阴森或过度泛黄。

画面下方约占六成，是现实中的父子。左侧是一位三十二岁左右的成年儿子，栗棕色略微凌乱的短发，五官柔和，穿灰蓝色卷袖衬衫，坐在靠近父亲的小凳上，身体稍稍向右前倾。他左手稳稳托着一只米白色陶碗，右手握着一把小勺，正准备送到父亲嘴边，勺子与嘴唇之间仍留有一点距离。他看着父亲，眼神专注、耐心，嘴角带着很轻的笑意。

右侧是一位七十八岁左右的年迈父亲，银白色侧梳短发，浓而柔软的白眉，面容清瘦，鼻梁略高，眼角有自然的皱纹。他穿燕麦色针织开衫和浅米色内衫，坐在扶手椅里，腿上盖着浅棕灰色格纹毯。双手自然放在毯子上，身体略微朝儿子靠近，准备接受喂食，神情安宁而亲近。不要表现得痛苦、濒危或过度虚弱。

画面上方约占四成，墙面上有一块边缘柔和的暖金色矩形光幕，像一段被光保存下来的童年记忆。它不是电视机，也不是下方现实人物投出的物理影子，而是一段独立存在的往事剪影。 光幕里面，幼年的儿子坐在左侧带小餐盘的儿童高脚椅上，约三岁，头发蓬松；年轻时的父亲坐在右侧，具有与下方老人相呼应的鼻梁和侧脸轮廓。他一手托碗，另一只手拿勺，正从右向左给孩子喂食，勺子同样停在即将接近嘴巴的位置。

回忆人物采用清晰可读的深暖褐色侧面剪影，轮廓略有柔光，能够明确辨认父亲、孩子、高脚椅、碗和勺子。下方是儿子从左向右喂父亲；上方是父亲从右向左喂孩子，形成方向相反、情感呼应的构图。 回忆剪影完整位于上方光幕内，不被下方人物的头部遮挡。

人物风格精致、温暖，五官略有动画式夸张，但身体结构自然可信；发丝、皮肤、针织纤维和毯子褶皱细腻。固定平视中景，父子以相向的四分之三侧面呈现，既看清表情，也看清喂食动作。下方人物与上方回忆都要清晰可读，不要用强烈景深把回忆虚化。情绪含蓄、真诚、不刻意煽情。

无文字、无字幕、无水印、无播放按钮、无进度条，无额外人物。

In this collection

3D scene of a son feeding his elderly father with a childhood memory on the wall — vertical 9:16 opening frame

A 3D render in animated-film style shows an adult son spoon-feeding his aging father in an armchair, while a glowing wall panel recalls the reversed childhood moment when the father fed him. Vertical 9:16 framing, designed as the first frame of a family short film.

Image prompts and examples

When to use this

Create a moving but restrained opening still for a short video about care passed between a father and son across two eras: one gesture, two directions, in a single fixed frame. The vertical format keeps both story layers readable on a phone screen.

How to recreate it

1. Specify a high-quality 3D animated-film look with gently stylized features, naturally believable bodies, and a fixed eye-level medium shot. 2. Give the lower two thirds to the present: on the left, a son around thirty-two with slightly messy chestnut hair, a grey-blue shirt with rolled sleeves, seated on a small stool and leaning right, one hand steadying an off-white ceramic bowl, the other holding a spoon that stops short of his father's lips, his gaze patient with a faint smile. On the right, a father around seventy-eight with silver hair combed sideways, thick soft white eyebrows, a thin face and faint wrinkles, wearing an oatmeal knit cardigan over a light beige shirt, a grey-brown plaid blanket over his legs, hands resting on it, body tilted toward his son, calm and attentive. 3. In the upper portion, place a warm golden rectangular light panel with soft edges on the wall, holding a childhood memory in clearly readable dark warm-brown silhouette: a toddler of about three in a high chair with a small tray on the left, the young father on the right, spoon in hand, feeding the child from right to left. Keep the silhouettes fully inside the panel, never overlapped by the heads below. 4. Light the room with late-afternoon honey-gold light, a few fine dust motes, and warm translucent shadows; warm brown walls, dark wood wainscoting, a caramel high-back armchair. 5. Build the mirrored action: the son feeds left to right below, the father fed right to left above. 6. Render fine hair, skin, knit, and blanket-fold detail, keeping both layers sharp with no depth-of-field blur on the memory. 7. Forbid all text, subtitles, watermarks, player UI, and extra characters.

Practical adjustments

Ask for silhouettes described as dark warm brown with slightly soft edges; without that wording, the memory can turn into an unreadable blur or a literal cast shadow.
State that the elderly father should look peaceful rather than frail or in distress; this is the main safeguard against melodramatic results.
If the model merges the two eras, reframe the memory as a self-contained luminous rectangle and note that it is neither a television nor a physical shadow.

Limitations

The image is generated with GPT Image 2.5 in a 9:16 aspect ratio; it is a still first frame and produces no motion. 3D generators often blend the two pairs of figures, clip the memory silhouette, or drift toward a dim, overly yellow room, so several attempts are usually needed. The shared nose and profile linking the two fathers remains approximate, and no specific result is guaranteed across runs.