Graduation Project: AI Video Autobiography ‘You’

What to keep from one person’s life

Older people like to talk, and their stories repeat. You, (〈當身,〉) is a capstone graduation project that started from a question: what if artificial intelligence listened to those stories? As the tagline puts it, “Apart, yet as my own body” (떨어져 있지만 내 몸과 같습니다). Students approached an elderly couple who walk the campus every day, built a relationship with them, and over four months of shooting turned it into one video autobiography.

Seven teams and twenty-one people took part: planning and directing, sound, AI, editing, VFX, cinematography, and design. As team lead, I coordinated the overall direction and production flow.

From interviews to data, from data to scenes

We placed AI wherever it was needed to bring the couple’s memories onto the screen.

• Collecting the stories — We prepared interview questionnaires with generative AI, transcribed the recorded interviews with speech recognition, and organized them as material for the script. We wrote the script with a language model in stages: picking out the core stories, then refining scenes and style.
• Visualizing the past — We turned past situations from the interviews into prompts and generated old photographs with image generation. This is the moment the interviews, which were audio material, became visual material. Video generation tools then gave the still images motion, and they became flashback scenes.
• Faces and voices — We created their younger faces with de-aging and face swap, trained a voice generation model on the subjects’ voices, and added narration. A lip-sync tool matched picture and sound.
• Changing the seasons — Shooting lasted four months, but we wanted to capture a year of the couple’s life. On photos of the same places, we used inpainting to add snow and remove green leaves, so the passing of the seasons shows on screen.

The same scene in autumn and in an AI-made winter. This is the device that made four months of shooting look like a year of walks.

For editing, an automatic cut-editing plugin produced the rough cut, and for color, the automation features of a grading tool set the tone. The seven songs on the soundtrack were also made with AI composition and arrangement, and the poster was generated as an image and then refined with Photoshop’s Generative Fill.

The conclusion we kept

When we finished, the team wrote down this conclusion: today’s AI has limits when it comes to making a complete film on its own, and that is why the human eye for using AI in the right place matters. This sense carried into Companion (‘동행’) and the work that came after it.