AI学员风采|康洁茹: 用AI重塑场景,以技术复活记忆

AI学员风采｜康洁茹：

用AI重塑场景，以技术复活记忆

AI Student Spotlight — Kang Jieru:

Reshaping Scenes with AI, Reviving Memories Through Technology

我是AI短片《非即时反馈》的导演康洁茹。这部短片最初的创作初衷是在快节奏的城市生活中，打捞那些被我们遗忘的、属于“家”的温度。如今的年轻人大多独自在外打拼，生活被工作填满。我希望能还原一个普通的下班夜晚，那种“身体很累，但灵魂渴望慰藉”的真实状态——在这里，做饭不再是为了果腹，而是一场自我疗愈的仪式。

I am Kang Jieru, director of the AI short film Non-Real-Time Feedback. The original inspiration for this film came from a desire to reclaim the oft-forgotten warmth of “home” within the rush of fast-paced city life. These days, most young people strive alone far from home, their lives consumed by work. I wanted to recreate an ordinary evening after work — that authentic state of being physically exhausted yet spiritually yearning for comfort — where cooking is no longer just about filling the stomach, but becomes a ritual of self-healing.

带着这份思考，我报名参加了北京电影学院动画AI研修班。如今课程结束，我想把这段宝贵的经历记录下来——既是给自己的总结，也希望能给同样在AI浪潮中探索的同学一些参考。

With this in mind, I signed up for the Beijing Film Academy’s AI Animation Workshop. Now that the course has concluded, I want to document this invaluable experience — both as a personal reflection and as a reference for fellow students navigating the AI wave.

北京电影学院动画AI研修班课程系统地介绍了主流的动画AI工具，除了工具操作，课程还涵盖了故事创作、分镜设计、角色设定、场景构建等动画创作全流程。老师带我们拆解了大量国内外优秀AI动画作品，分析它们“哪里成功、哪里还有提升空间”。

The Animation AI Workshop at Beijing Film Academy systematically introduces mainstream animation AI tools. In addition to tool operation, the course covers the entire animation creation process, including story development, storyboard design, character setting, and scene construction. The instructor guided us through a detailed analysis of numerous outstanding AI-generated animations from both domestic and international sources, examining what works well in them and where there is still room for improvement.

   通过学习让我意识到：AI动画的价值，不在于像素对现实的穷尽模仿，而在于能否用技术为记忆塑形，让消逝的时光再次被看见、被触碰。

Through this learning, I came to realize that the value of AI animation does not lie in the exhaustive imitation of reality through pixels. Instead, it lies in the ability to use technology to shape memory — to let fading moments be seen and felt once again.

在研修班的学习过程中，老师对短片一些关键部分的表达给与了指导。

During the workshop, my instructors provided valuable guidance on expressing key parts of the short film.

在指导老师的帮助下，我学会了如何让提示词更精准，人物要有表情，画面要有结构。正是带着这些方法，我完成了对“用AI重塑场景，以技术复活记忆”这句话的第一次完整实践。这不仅仅是一个视频，更是一部关于成长、怀念与自我重建的短片。它想告诉每一个异乡人：无论走多远，胃记得回家的路。而我作为创作者，知道那些记忆长什么样、有什么味道、该被如何讲述。我认为AI不是“对手”，而是“记忆的修复师”。

With the help of my instructors, I learned how to craft more precise prompts — ensuring characters expressed emotion and scenes had structure. Armed with these methods, I completed my first full realization of the idea: “reshaping scenes with AI, reviving memories through technology.” This is not merely a video; it is a short film about growth, remembrance, and self-reconstruction. It tells every sojourner: No matter how far you roam, your stomach remembers the way home. And as a creator, I know what those memories look like, what they taste like, and how they deserve to be told. I believe AI is not an “opponent” but a “memory restorer.”

红烧肉在这里不只是一道菜，它是连接过去与现在的开关。小时候，红烧肉代表无忧无虑，代表姥姥的爱——那种小火慢炖的耐心，是现代快餐永远无法替代的。长大后，复刻这道菜，是试图在陌生的城市里，重建童年的安全感。当酱香弥漫厨房的那一刻，时空折叠了。

Here, braised pork is more than just a dish: it is a switch connecting past and present. As a child, braised pork meant carefree happiness and grandmother’s love — that slow, simmering patience that fast food today can never replace. As an adult, recreating this dish is an attempt to rebuild the security of childhood in an unfamiliar city. The moment the rich fragrance fills the kitchen, time folds in on itself.

在短片创作中，我用AI辅助生成了老式厨房的灶台、窗格、蒸汽等场景细节。这些画面并非完全还原我的童年，而是用技术“重塑”了一个承载记忆的容器——它足够真实，让观众想起自己的家；又足够温柔，让记忆不至于被像素“冻住”。

我们不常说爱，但会把爱炖进肉里，咸甜适口，余味悠长，该视频表达了一种含蓄的中国式亲情。视频结尾回到现实，不是为了表现伤感，而是为了表现“吃饱了，就有力量继续面对明天”。姥姥虽然不在身边，但那份爱已经变成了女孩手里的锅铲，支撑她好好生活。

In creating the short film, I used AI to help generate scene details: the old-fashioned kitchen stove, the window lattice, the steam. These images do not precisely restore my childhood — rather, technology “reshapes” a vessel that carries memory. It is real enough to remind viewers of their own homes, yet gentle enough not to freeze memory cold in pixels.       We seldom speak love aloud, but we braise love into the meat — savory with a touch of sweet, a lingering aftertaste. The film expresses this subtle, Chinese-style familial affection. The ending returns to the present not to evoke sadness, but to convey: “When you’re full, you have the strength to face tomorrow.” Grandmother may no longer be by her side, but her love has become the spatula in the girl’s hand, supporting her as she lives on.

这部短片如果纯手绘制作，预估周期需要数月。借助AI辅助生成场景和部分素材，我能够把更多精力放在故事线、情绪节奏和关键原画上——这才是短片真正打动人的地方。效率的提升，让我有余力把情感打磨得更细腻。

以前画一个复杂场景的背景可能要一两天，现在AI可以快速生成多个版本。但对我来说，更重要的是：当我脑海中的厨房记忆越来越模糊时，AI帮我生成了那个灶台、那口砂锅、那缕蒸汽——然后我对照着记忆，一点一点把它们调整成“对的”样子。这个过程，像是在用技术修复一张老照片。不是凭空创造，而是让模糊的变得清晰，让遥远的变得触手可及。

If this short film had been made entirely by hand, the estimated production cycle would have been several months. With AI assistance in generating scenes and certain assets, I was able to devote more energy to the story, emotional pacing, and key original drawings — which is what truly moves the audience. The gain in efficiency gave me the leeway to polish the emotions more delicately.      In the past, painting a complex background might take a day or two. Now AI can quickly generate multiple versions. But what matters more to me is this: when the kitchen memories in my mind grew increasingly hazy, AI helped generate that stove, that clay pot, those wisps of steam — and then, comparing each element with my memories, I adjusted them little by little until they felt “right.” This process is like restoring an old photograph with technology. Not creating out of thin air, but making the blurry clear and the distant reachable.

在短片创作过程中，我完整经历了从构思到成片的各个阶段，有成功也有失败。AI生成素材中，我遇到了两个棘手的问题：

During the short film creation process, I went through every stage from concept to final output, experiencing both successes and failures. While generating source material with AI, I encountered two thorny problems.

我希望女孩脸上有“疲惫被治愈”的复杂表情，但AI生成的结果要么太夸张像在演戏，要么太呆板毫无情绪。尝试了十几个版本，没有一张能用。在老师指导下，我放弃了让AI“一步到位”的想法，改用AI生成基础面部结构，再在后期软件中手动调整眉眼和嘴角的弧度。

I wanted the girl to have a complex expression of being "exhausted yet healed." But the AI-generated results were either too exaggerated, as if she were acting, or too dull and emotionless. I tried over a dozen versions, and not a single one was usable. Under my instructor's guidance, I gave up on the idea of having AI "get it right in one go." Instead, I used AI to generate the basic facial structure, then manually adjusted the curves of the eyebrows, eyes, and mouth in post-production software.

不同镜头里的灶台颜色、光线方向、锅盖款式都对不上，拼在一起像穿帮的拼贴画，根本没办法剪辑。经过老师的指导，我建立了一套核心提示词模板，每次生成时都保留“暖黄夕阳光从左侧照入”“黑色土砂锅”“木质锅盖”等关键词，并锁定同一张视觉风格参考图。场景终于统一了，镜头之间可以顺畅衔接。

The stove colors, lighting directions, and pot lid styles in different shots didn't match. When pieced together, they looked like a collage of continuity errors, completely uneditable. With my instructor's help, I established a core prompt template. Every time I generated an image, I kept keywords such as "warm yellow sunset light entering from the left," "black clay pot," and "wooden lid," and locked in the same visual style reference image. Finally, the scenes became consistent, and the shots could be smoothly connected.

经过老师的指导，这些问题都得到了成功解决。

With my instructor's guidance, all of these problems were successfully resolved.

素材备齐后，最关键的问题是：这些镜头该怎么剪？

我的核心思维是：用节奏的“卡顿”反衬情绪的“流畅”。

After all the footage was ready, the most critical question was: how should these shots be edited? My core directorial thinking was: use rhythmic "stuttering" to contrast with emotional "fluency."

前半段表现疲惫时，我用了快切蒙太奇——手机消息、倒计时闪烁、划动短视频的手指，节奏快且令人烦躁，让观众感受“信息过载”。

后半段做饭开始后，剪辑瞬间慢了下来。洗肉、切肉、炒糖色、盖上锅盖，每个动作都给足时长，蒸汽镜头甚至用了慢放。这种节奏的突然放缓，是为了让观众“被迫”慢下来，像女孩一样在等待中找回呼吸的节奏。

声音也做了刻意设计：前半段是紧张刺耳的手机提示音、电子“叮”声；后半段是舒缓的砂锅“咕嘟”声、油“滋啦”声、菜刀“笃笃”声。声音的变化比画面更直接地告诉观众：你进入了一个不同的时空。

In the first half, to convey exhaustion, I used fast-cut montage — phone notifications, flashing countdowns, fingers swiping through short videos. The rhythm was fast and even irritating, making the audience feel the same "information overload."

In the second half, once the cooking began, the editing instantly slowed down. Each action — washing the meat, cutting the meat, caramelizing the sugar, putting on the lid — was given ample duration. The steam shot was even played in slow motion. This sudden deceleration was a deliberate choice: to force the audience to slow down, just like the girl, and find their breathing rhythm again in the waiting.

The sound was also deliberately designed. The first half featured tense, piercing phone notification sounds and electronic "dings." The second half featured soothing sounds: the "gurgle" of the clay pot, the "sizzle" of oil, and the "thump-thump" of the knife on the cutting board. The change in sound tells the audience more directly than the visuals: you have entered a different time and space.

从北京电影学院研修班回来后，我最大的感受是：技术能唤醒我们内心的记忆，推动我们成为更好的自己。

未来的动画行业，一定是“人的记忆 + AI的技术”的时代。而我们需要做的，是让自己的情感和审美，配得上这些工具。

Returning from the Beijing Film Academy workshop, my greatest realization is this: technology can awaken the memories within us and propel us to become better versions of ourselves. The future of animation will surely be an era of “human memory + AI technology.” And what we need to do is to ensure our emotions and aesthetic sensibilities are worthy of these tools.