When I first saw the AI video, Kling had just been released, and I was immediately captivated by it. I made a 5-minute Batman video. Although this video has become outdated and unwatchable due to technological advances, I got sick from overexertion in my first week of making AI videos. I had to lie down for several days. That’s how fascinated I was.
Since I started keeping records, I’ve spent 1,690 hours on AI video. While I’m still far from the 10,000-hour mark, this has long surpassed being just a hobby. I became obsessed with AI video to visualize my novels. I write original science fiction, and sci-fi is the most suitable genre for AI expression.
AI can achieve specific special effects at a low cost, and at least in terms of atmosphere, AI can fully meet the requirements of science fiction films. Additionally, AI isn’t good at expressing human emotions, so it’s not well-suited for modern or historical romance. Science fiction typically focuses on showcasing concepts and presenting sci-fi ideas, so combined with AI’s grasp of special effects, it’s like adding wings to a tiger.
Driven by my initial desire to visualize my novels, I embarked on my AI film production journey. Before making AI videos, I had been a programmer for over ten years, so I always approached it from a technical perspective. Many people believe that technology isn’t key—art is—so they resist technical operations like training LoRAs or using ComfyUI. But this poses no difficulty for me.
In the second half of last year, I spent six months researching and testing technology. Though new technologies arrive every month, the shelf life of learned techniques is only a few months, sometimes even one month. But I have my obsessions: I persistently pursue character clothing and consistency and the atmospheric feel of scenes. Therefore, I use a hybrid approach combining Flux and Midjourney for image generation.
Many believe that current AI can’t maintain stable consistency, so we shouldn’t pursue character consistency or narrative. They either think people won’t care much about consistency, or they consider ComfyUI a difficult tool that will be replaced. Consistency is vital because it’s called silent storytelling.
Silent storytelling typically occurs in games, where the game environment, notes within it, and character clothing all tell stories. What characters say is less important than what remains silent. People keenly absorb many visual and auditory details, so I pursue character consistency.
I want my designated characters, wearing my designated gorgeous or minimalist clothing, to perform my designated scenes in my designated settings. Of course, as technology advances, character consistency will definitely be resolved appropriately. I believe in the purity of artistic expression and the beauty of technology.
After researching AI production technology, I studied directing techniques. Directing is also a form of technical learning, with audiovisual language as the foundation. In October last year, I made a series of AI videos that could only be described as passionate but were completely chaotic. The result was that they looked very amateurish.
In filmmaking, the first key point is still consistency—the overall style must remain consistent across every shot. This consistency can be said to be the source of art. When we use AI image generation tools, there’s consistency within each tool—for example, Midjourney has its own style. But once we cross tools, inconsistency appears. This inconsistency can mostly be resolved through color grading.
But the most important aspect is the overall tone—whether the content in the frame belongs to the same world. I initially found this difficult to grasp; passion alone cannot achieve this—it requires discipline. That is, the director should design everything appearing in each shot. This design sensibility ensures tonal consistency.
When color grading, I prefer a darker style with low saturation. This has been criticized as too American, but I still stick to my artistic expression. We’re already in the AI film era, and I believe the creator should decide what stories to tell and how to tell them.
In the second half of 2025, I will gradually visualize my novels. Starting with short stories, I’ll make short films based on them. I’ve prepared 25 concepts similar to “Love, Death & Robots.” They have a dark, absurd feeling within the Computational Universe with intricate connections. I’ll publish short stories on my blog when the short films go online.
Next year, after my novel is published on Amazon KDP, I’ll start adapting it into a series and updating it simultaneously. This relates to the current characteristics of AI technology—it updates every month, and films made with old technology quickly become unwatchable. Therefore, I adopt the strategy of making short films first and doing feature-length work when the technology matures.
I am an author, but I became a director to let everyone intuitively experience the Computational Universe. This is the cross-disciplinary nature of the AI era. It’s beautiful because AI art will be the tenth art form, combining technology, art, and literature.
Welcome to the computational universe, world travelers!
