This involves several issues, such as AI’s current limitations in long-form narrative, AI performances lacking emotional depth, and audiences having high overall expectations for films. We are amazed by AI’s development on one hand, while lamenting that it’s still too far from being watchable on the other.
Although I am an all-in supporter of AI film and television, I also have doubts. Sometimes I wonder whether this promise of “one person being an entire film crew” is real, and whether it can truly move audiences in long-form narratives. Some people say a good story is the foundation for moving audiences, but what if, no matter how hard we try, we cannot achieve this due to AI limitations?
AI films started with a very PowerPoint-like feel last year, but over the course of one year, the technology has developed by leaps and bounds. By May or June of this year, new technologies will have enabled every creator to avoid the spread of the PowerPoint feel. Now, multi-parameter capabilities can further control character consistency.
Style consistency, character consistency, and scene consistency – these three consistencies are the foundation of long-form narrative. Previously, making slightly longer films, such as ten minutes, would have been difficult. Audiences might only be able to watch the first 3 minutes before losing interest. That’s why popular AI videos abroad are not long.
The direction I’ve been researching is stable long-form narrative. During my research, I discovered that, in addition to the three consistencies, what also troubles me is shot design and overall pacing. I often feel like I’m just editing a music video rather than making a short film. It’s possible that my abilities in film narrative are insufficient, or it’s possible that current AI has only evolved from PowerPoint to music videos.
As for AI performances lacking emotion, I’ve heard people say more than once, “It’s made by AI, it has no soul, I won’t watch it.” Or upon hearing it’s AI-made, they don’t want to watch anymore. AI-driven systems seem to be cheaply labeled as “soulless.” Every time I see an AI face that’s absolutely symmetrical and perfect, I don’t have much desire to click on the video either. Although I watch peers’ work to gauge everyone’s level.
Currently, most audiences for AI films are also peers—curious explorers and early adopters. To reach mass adoption, we still need to pass the emotional barrier. Emotion isn’t necessarily just character expressions, but character expressions are part of it. Veo3 has made great breakthroughs in this area, but Kling can compete with it in image-to-video generation.
We should ask ourselves: if I were an audience member, would I watch AI movies? Current AI videos are of mixed quality, and I only watch them within professional contexts – they haven’t become entertainment necessities for me. My entertainment necessities are still superhero movies, which have hundreds of millions of dollars invested in them.
Today’s audiences have long since watched big special effects movies. While AI-generated films amaze creators who think, “This can actually be filmed,” for audiences, there’s no novelty – they might even find the uncanny valley effect severe. Especially for realistic-style AI films, none currently avoid the uncanny valley.
So is this just industry self-congratulation? First, AI is definitely not a bubble, not as some say, purely media hype. However, the quality that current AI films can achieve in coherent, long-form narratives indeed has various shortcomings. They’re quite good for making trailers with atmosphere. We even see some accounts that only make trailers without complete narratives.
Some believe AI movies won’t have real audiences within ten years.
Since we can’t achieve optimal results now, should we not do it? Many people who have dabbled in AI film and television have actually given up or are waiting and watching. As creators, each film release allows us to gain certain experience. So, I’m in a half-waiting state – I’ll make short films first while accumulating material for long-form novels. I might not start filming features for another 2 years.
My eight-part, long-form novel, “The Computing Universe,” features emotionally rich and detailed action scenes. Although Conch 02 already has some capability for filming action scenes, action scenes with strong emotions require too many details to perfect. This level of detail is beyond current technology’s reach. Additionally, I believe “The Computing Universe” is too important for me and for the computing universe itself – I need to time the technology correctly.
Regardless of whether large audiences watch, I won’t forget my amazement when I first saw Kling and immediately created a Batman AI video. I won’t stop doing this just because I’m not achieving optimal results now. Although there may only be professional audiences, each of my films is better than the previous one. It won’t take ten years – perhaps just two years for AI technology to reach maturity.
Follow me and explore AI filmmaking together.
