It’s June 2025, and AIGC filmmaking has been developing for one year. This industry is entirely new and far from being fully established, though leading players have emerged. The future can accommodate many more creators because it’s easy to start. Whether professional or amateur enthusiasts, they will gradually enter this industry.
AIGC film and television have moved beyond the initial pure PowerPoint-style presentation and entered an era of free camera movements and diverse styles. However, they still have many limitations, causing creators to often feel both frustrated by their constraints and delighted by occasional flashes of creative brilliance during production.
Many people say AI-generated content lacks soul, which is the first limitation of AI filmmaking. Characters have an AI-like quality – perhaps too perfect, maybe too stiff. In any case, the uncanny valley effect lurks subtly, especially with realistic human styles. Not everyone can accept this.
AI tools currently lack souls, but the people using them have souls. Artistic expression fundamentally manifests the soul, regardless of whether AI is used. AI film quality will only improve as tools become more advanced. I think we can also avoid categories that particularly require human emotion and focus on what AI is naturally better at, such as science fiction.
AI can handle grand science fiction scenes and the stylization that science fiction requires. Although Chinese science fiction literature and film have never been particularly popular, a new era has arrived. We don’t need to educate audiences on how to appreciate science fiction; we need to express and liberate our original creativity. Fantasy and supernatural genres are the same. AI can leverage its strengths when working with fantastical themes.
In the future, AI’s soul might manifest itself in its unique new aesthetic style. Artistic expressions that previous tools couldn’t achieve might become distinctive features of AI art, though I’m not talking about that pronounced Midjourney-generated image flavor here.
When I first encountered AI video tools, I fell in love with AI because I saw not its flaws but infinite possibilities. This might also be because I’m not an artist, so I don’t feel that my profession is being violated. When I see AI writing novels, I feel somewhat offended because of my years of writing experience—but I still learned how to use AI for novel writing, even though I won’t use AI to write novels myself.
AI will never gain everyone’s support. I know many creators are very resistant to AI. First, because image generation AI used artworks published online by artists for training without obtaining permission from all artists and didn’t share corresponding benefits with them.
Second, when AI art first appeared, the quality wasn’t optimal. AI artists “polluted” the original ecosystem with AI-generated images, making many artists feel disgusted and launching anti-AI art movements. Not just artists but many other people online also attack anything AI-related.
I’ve been attacked myself, but I still believe in AI. AI is the ultimate expression of art, technology, and literature, with benefits far outweighing its flaws. It enables everyone to become a director. Not only that, everyone can potentially become some kind of artist. This is unprecedented creative democratization.
Now, let me discuss some more of AIGC’s operational limitations. For example, it is challenging to maintain scene consistency. Using Midjourney’s style references can achieve some stylistic consistency in scenes, but the spatial sense of scenes still cannot be entirely consistent. Dreamina 3.0’s multi-storyboard feature can achieve a certain level of consistency that is stronger than Midjourney’s. In the future, multi-reference videos should also help solve scene consistency issues or integrate with 3D tools.
AI’s performance in action sequences also needs improvement. Current AI tools aren’t incapable of creating action scenes – it’s just that the operational difficulty is high, and the cinematic presentation hasn’t reached a certain standard. The foundation of action sequences is fight scenes between two people, which hasn’t been fully resolved yet. My “Computing Universe” series is filled with nuanced action sequences with character emotions, which currently cannot be produced. Therefore, the visualization of “The Computing Universe” novel must wait for technological maturity. Before then, I’ll work on short stories for testing.
AI also faces particular difficulties in long-form narrative storytelling. We’ve seen many short films produced with good results. However, once a long-term narrative is involved, audiences easily become fatigued. Shots still can’t provide much variation, and the PowerPoint feeling returns once the duration is extended. I originally planned to adapt “The Computing Wandering” into a long series with 10-minute episodes totaling 240 minutes. But later, I discovered that getting audiences through even the first 10 minutes was problematic.
AI technology is developing rapidly. Let us look forward to the future, which is already here.
