Captured, Not Generated.
- Jul 10
- 2 min read

I recently rewatched Goodbye South, Goodbye by Hou Hsiao-Hsien, one of my favourite filmmakers and someone who has inspired some of my own films.
I still revisit movies I love every now and then whenever I have the time. Even though I already know the story, it's always the little details I've forgotten over the years that surprise me again. Sometimes I even come away with a completely different perspective. I think the way we watch and interpret films changes as we grow older and collect more life experiences. A scene that felt ordinary when I was younger can suddenly carry an emotional weight that I never noticed before. It's almost like watching a different film.
Rewatching Goodbye South, Goodbye reminded me why films like this are so important. It also made me think about something that's surprisingly difficult to replicate with AI.
It's not that AI can't generate images like these. It absolutely can. It can recreate the mood, the atmosphere, the lighting, the camera movement, even produce spectacular visuals and Hollywood-level effects if you prompt it well enough.
But this film wasn't generated. It was captured.
It was filmed at a specific time, in specific places, with people who lived through that particular moment in history. Those images weren't assembled from billions of data points. They were recorded as fragments of reality. The actors occupied those spaces. The streets existed as they were. The director was expressing what he felt and observed at that point in his life.
Could we make something similar with AI in the future? Maybe. We can certainly imitate its appearance. But we all know it wouldn't be the same thing.
That also made me realize something about my own process.
There were several beautiful scenes that stood out to me again during this rewatch. Of course, I could probably recreate shots that look very similar with AI today. But then another thought occurred to me: if I had never watched this film in the first place, would I even know how to frame those shots? Would I know how the characters should be blocked, or why the camera was placed where it was? How would I even know what prompt to write?
The prompt doesn't come first. The experience does.
That's why I think these films, the ones often labelled as "arthouse," have become even more valuable in the age of AI. AI can learn patterns and generate images, but it doesn't understand why a particular image was worth making in the first place. Those reasons come from people, from memory, from history, from observation, and from living.
Maybe that's why I still keep returning to these films.
Maybe that's also a sign I'm getting older.
Comments