After completing this week’s work, next week I’ll start producing “The Withered Sea Project.” This novel adaptation is being prepared for this year’s Vacat Award, and it’s more suitable than my previous work, “The Resolve,” because this is a Chinese-style science fiction story. At the same time, that one was a steampunk story. “The Withered Sea Project” takes place in the Xinglong Civilization, which represents China in the computational universe.
The story is set in the year 2700 of the Xinglong Era. After the victory of the Fourth Stellar War, mathematician Yu Baiyun feels deeply concerned about the future of civilization. She focuses on collective cognitive model research and initiates the Withered Sea Project for simulation experiments, but this leads her into a mental crisis.
Although her warnings spark controversy and even lead to her exile, her research is inherited by her student Liang Moqing. Amid the clamor of victory, Yu Baiyun’s reflections on civilization’s spirit of exploration become a crucial question hanging over Xinglong’s future.
My first instinct is that this will be challenging, but I’m having trouble pinpointing exactly where my concerns lie. Before officially starting, let me analyze this vague worry.
First and foremost, this will be challenging because the story is almost entirely dialogue-driven. I’m having trouble finding a suitable place to insert action scenes. In short, it’s dull. While purely dialogue-driven live-action films can avoid being boring, when AI does it, it easily becomes like a PowerPoint presentation.
How does it become PowerPoint-like? It tends to feature a lot of character dialogue, and AI currently requires significant adjustment even to create a good shot-reverse shot sequence. For example, when using AI to develop over-the-shoulder shots, sometimes Photoshop is needed because it involves two characters in the same frame.
Beginners handling dialogue scenes often only use medium shots, with one shot per character’s line of dialogue. I used to do this, too. Only after seeing more professional approaches did I learn about over-the-shoulder shots, which don’t always focus on the speaker, and the importance of avoiding constant medium shots. Yet this still easily becomes PowerPoint-like.
Scenes that consist entirely of dialogue have a classical feel to them. For instance, I find it hard to accept the new “The Sandman 2,” which is almost entirely dialogue from start to finish. Using AI to create this would be even more disastrous.
Secondly, this novel’s concepts are overly conceptual – written for the sake of science fiction ideas rather than for the story itself. After I finished writing it, I didn’t even know how to approach the script. Later, I used Claude to adapt it into a screenplay, then addressed all the problems that AI writing typically has, such as awkward dialogue.
Claude can indeed break down complex novels into simple scripts. Then, yesterday, following a group member’s suggestion, Minimax’s Agent can also create very detailed shot lists for screenwriting. These two AI tools have helped ground this rather abstract novel.
AI filmmaking naturally has advantages in big-budget science fiction spectacles, but pure dialogue scenes and non-spectacle conceptual content can diminish these advantages. Compared to current live-action short dramas, AI would have no cost or time advantages when doing the same subject matter. The advantage lies in doing science fiction with special effects that live-action short dramas can’t achieve.
The third challenge lies in this novel’s artistic style. In “The Resolve,” the artistic style was clear and thoroughly developed, characterized by a steampunk aesthetic. But “The Withered Sea Project” might be some kind of futuristic Chinese style? Even Midjourney might not fully understand this.
Using a thoroughly developed artistic style helps maintain consistency across multiple tools, enabling Midjourney to generate style-consistent scene images. These scene images can be used as reference images or input into other tools. Therefore, Midjourney is currently the primary source of AI-generated aesthetics.
I have 25 short story ideas that I plan to write into short stories one by one, then film as AI short films. These 25 stories encompass a range of high-concept pieces and pure dialogue-driven narratives, each presenting its own unique filming challenges. However, I’ll continue to explore the boundaries of what AI filmmaking can do, so I’ll make them one film at a time.
Future plans involve creating short stories and films while continuing to write novels and exploring opportunities to produce longer films. However, all content will be set within the Computational Universe, ensuring continuity and consistency. In the computational universe, there are various civilizations composed of different alien species. Can aliens achieve good consistency? Humanoid aliens should be possible, but what about non-humanoid ones?
“The Withered Sea Project” starts next week – another battle between film and PowerPoint.
Follow me as we explore AI filmmaking together.
