How did a non-professional filmmaker’s AI project stand out among ordinary short films?
In February 2023, Jim, a programmer in the gaming industry, became an AI artist.
By December 2024, Jim’s AI animated short film “The Thin Man The Gun The Hotpot” won the Best Animated Short Gold Award at the Independent Shorts Awards.
This is a significant international short film festival in the independent film world, not specifically for AI works, making it a major achievement for Jim. He stated, “It’s gratifying that the judges focused on the story itself, ignoring the AI label.”
“The Thin Man The Gun The Hotpot” was primarily completed by Jim alone. The 10-minute film consists of 242 shots, took nearly three months to make, and required at least 200 hours of work, averaging over two hours a day.
During the 90-day creation process, Jim felt like he was in a battle of wits with AI, pushing its limits and avoiding its constraints. When viewers remarked, “AI has started making animations,” Jim responded, “It’s the creators driving AI to make animations.”
3 Months, 10 Minutes
“The Thin Man The Gun The Hotpot” is a noir crime film, a significant genre in Hollywood’s past, with a story rooted in Chinese culture. The term “The Thin Man The Gun The Hotpot” comes from Southwest China, referring to small food stalls open late at night.
The protagonist, Xu Xia, is a young man eating at a street stall late at night. To pay for his father’s medical treatment, he makes unethical choices but maintains an untouchable moral line. Ultimately, he becomes entangled in violence and murder, trapped by fate.
Rather than calling it “AI animation,” it’s more accurate to say “The Thin Man The Gun The Hotpot” is an animation made using AI tools.
AI generated the visuals, while other elements like the script, editing, voice acting, music, and sound effects were done manually. The text in the visuals was added in post-production.
In terms of visuals, Jim follows a “pure AI generation” style. “The Thin Man The Gun The Hotpot” has no live-action footage; it relies on image-to-video conversion. Images were generated by Midjourney, and videos by Keling, Pika, Jidream, PixVerse, and Runway.
AI generation is unpredictable, but telling a coherent story with AI requires stability. From character design, Jim considered how to maintain character consistency.
Jim had two principles for character design. First, the overall appearance should be simple, describable with a few keywords. Second, characters should have distinctive elements, so even if not entirely consistent, they remain recognizable to the audience.
The most representative characters are Brother Wan and Li Jiajia. Brother Wan is bald, wears sportswear, and sunglasses; Li Jiajia resembles a retro 90s woman, wearing red with wavy hair.
The seemingly refined Mr. Zhu, wearing glasses and a suit without a beard, lacks any unusual features, making him the most iterated character by Jim.
Jim found that AI easily generates characters with an “old money” aura, but struggles with characters like Mr. Zhu, who is wealthy but not elite, dangerous but not a crime boss.
“Live-action film with an animation filter,” is how Jim describes his animation style. Viewers can imagine what it would look like with real actors.
Once each character’s image is generated through text, it’s like having a movie’s costume photo. Jim then uses these images as material, generating more images with different angles and scenes while keeping the characters unchanged.
With enough images, videos can be generated. AI video tools update quickly; “The Thin Man The Gun The Hotpot” used versions from August to September 2024.
For technically challenging shots, Jim experimented with various video tools. Each tool has its strengths. Jidream, Keling, and Pika were the most used tools.
At the time, Jidream excelled at action shots, Pika at scene shots and simple talking animations, while Keling had strong overall capabilities, though sometimes unnecessarily complex in specific situations.
Even with today’s tools, many scenes in “The Thin Man The Gun The Hotpot” are challenging to achieve naturally. This requires relying on traditional methods.
For instance, when characters are not facing directly forward or in close-ups that AI can easily recognize, and when actions, expressions, and camera angles are in motion, a voice actor must dub according to the video’s lip movements. Jim believes that new features aren’t necessarily useful; old methods, though perhaps clumsy, are more reliable.
Niche and Diverse
The story of “The Thin Man The Gun The Hotpot” is set in a small town in southwest China, with characters speaking Sichuan dialect, a personal preference of Jim’s. The art style of “The Thin Man The Gun The Hotpot” is also unique, with viewers comparing it to an episode of “Love, Death & Robots.”
In his work in the gaming industry, Jim has seen too much similar, tried-and-tested content.
AI-based creation can be personalized. High-quality works still require large companies to invest significant resources, but Jim sees more possibilities in creating “a basic work.”
From the start of using Midjourney, Jim was clear that he didn’t want to create in a “mainstream” style. As a heavy user of Midjourney, he is well aware of the styles AI easily generates, which are also his no-go zones.
He prefers styles that aren’t immediately recognizable, with a retro feel but not entirely old-fashioned animation—familiar yet novel.
For the art style of “The Thin Man The Gun The Hotpot,” Jim referenced animation director Masaaki Yuasa and manga artist Yoshiharu Tsuge.
He first used Midjourney’s Describe feature to understand the keywords of their styles, then continuously wrote prompts, iterating repeatedly to generate images that met his requirements.
The worldview of “The Thin Man The Gun The Hotpot” largely stems from the regions and elements Jim is familiar with.
In 2019, Jim watched “The Wild Goose Lake” directed by Diao Yinan and starring Hu Ge, and was deeply impressed by its “down-to-earth setting with a noir aesthetic and core.” He realized that traditional genre films are not set in stone, and there is new vitality in Chinese narratives.
In 2021, Jim began writing the script, forming the basic outline of “The Thin Man The Gun The Hotpot,” including a small town, a desperate young man, and a life predicament.
In 2023, Jim started using Midjourney to learn AI concept art.
However, before 2024, Jim never thought about actually making an AI film. He taught himself photography, scriptwriting, and cinematography but stopped there. Videos tell stories through lenses, but he couldn’t create his own set shots.
In early 2024, Jim started using some of the earliest AI videos like Pika and suddenly realized he could finally animate scenes, adjust and modify images, and connect them to complete a short film. His desire to create became unprecedentedly strong.
With the materials ready, it was time for editing, sound effects, dubbing, and music. Jim delved deeper into the traditional film production process. He experimented and learned, gradually uploading some stylized, personally distinctive works on Bilibili.
Jim’s other AI short film, “Hard Bop Gunman,” drew from the style of old American comics and silent films, later winning an award in the Creative Track of an AI video competition on platforms like 1905 Movie Network and Bilibili.
AI has yet to disrupt many industries, but for small teams and individual creators, the cost of creation has become acceptable, and diverse content has a chance to be seen by audiences.
Setting “The Thin Man The Gun The Hotpot” in the Sichuan-Chongqing region, Jim admits he was somewhat “deliberate.” He is proving that when creating niche content, there’s no longer a need to worry too much about the traditional notion of “too risky.”
Limited, Yet Unique
A 10-minute “The Thin Man The Gun The Hotpot” might not be considered a long video on Bilibili. But in the AI community, 10 minutes is a rare length, not to mention that “The Thin Man The Gun The Hotpot” tells a complete story.
To complete these 10 minutes, Jim pushed both himself and AI to the limit.
The action scenes in the hotpot restaurant in “The Thin Man The Gun The Hotpot” took a lot of effort from Jim, with each shot designed, but the final effect presented less than 40%.
Jim admits that AI’s action scenes are indeed lacking, with no sense of impact and not following physical logic, like hitting air. Not just action scenes, but any shot with significant movement, like choking, throwing a hammer, or smashing with a stone, is difficult for AI to achieve.
According to Jim’s experience, if you must design action scenes for AI, avoid cold weapons and use firearms instead, as AI can at least simulate shooting. He himself used this technique, “Thanks to old Hong Kong films for the rescue.”
When real actors struggle with action scenes, stunt doubles can step in, but expressive facial expressions are an actor’s basic skill. However, this is also a weak point for AI, which either overacts or shows no expression.
In October 2024, Runway launched the Act-One feature, which drives AI characters to make the same expressions based on real human performance videos. Jim sees this as a sign that if more products launch similar features, it indicates that AI indeed has a bottleneck in performance.
In “The Thin Man The Gun The Hotpot,” characters often appear in close-ups and half-body shots, reflecting an AI weakness: handling shots with multiple subjects. When designing shots, Jim tries to avoid scenes with multiple people because he hasn’t found a good solution yet.
Even when generating images with Midjourney, if there are more than two people, facial issues may arise. When placed in AI video interactions, the scene becomes even more chaotic.
AI-generated visuals have many limitations, so in post-production, Jim uses close-ups, symbolic shots, and quick cuts to compensate for visual shortcomings. The recurring animal imagery in “The Thin Man The Gun The Hotpot” is an example.
Metaphors and symbolism are also commonly used in some classic American noir films. In the 1940s and 50s, due to technical limitations and regulations like the Hays Code restricting violent imagery, the situation was somewhat similar to AI videos in 2024.
The echo across decades is intriguing to Jim, “Perhaps outdated methods work well in AI films.”
However, audiences are honest, offering feedback like “a smooth PowerPoint presentation” or “an enhanced dynamic comic.” Jim accepts this, which is why he chose to make AI animations.
Compared to the smooth-skinned, eerily moving real-life style, animation at least “doesn’t make viewers immediately think, wow, this is too fake.” The slightly stiff and rigid movements of characters are more acceptable in non-realistic animation.
Not blindly pursuing length and ensuring quality first is partly because Jim benchmarks against regular films. He entered the non-AI competition Independent Shorts Awards to make judges overlook the AI label and focus on the story itself.
Jim hopes that when watching his work, audiences can be aware of AI’s presence but not come for AI, not showing leniency in the AI film evaluation system, “The AI community is too friendly, hearing some harsh words is necessary.”
Persisting in Creation to Overcome Anxiety
AI news often writes about “overnight changes,” and creating a short film in three months is already a long time.
AI progresses daily. In 2023, Jim could keep up with daily AI information, but from early 2024, keeping up with trends became impossible. He also feels anxious but not blindly so, like many others.
Technology is powerful, but there is still much work for humans to do. The difficulty and interest in creation lie in the fact that some problems must be solved by the creator themselves.
In “The Wild Goose Lake,” there’s a gunfight scene that left a deep impression on Jim: people are dancing in a square, plainclothes police wear glowing shoes, criminals are shooting, police chase with glowing blood, and then shoot too.
The tense yet playful contrast moved Jim. Without this scene, the first reaction to square dancing would be “tacky.”
He believes that often, familiar elements are not outdated but lack deep exploration by creators.
Conversely, things carefully expressed with old tools may not be replaced by new technology.
AI videos have been updated many times, but Jim doesn’t plan to remake previous AI shorts.
The design and compromises of the entire short film were based on AI’s limitations at the time, forming a complete entity. Even years later, there are still fun aspects, which is where the value of the work lies.
AI can help Jim complete more similar experiments, rethinking interesting things. He doesn’t like AI directly providing answers; he prefers solving problems himself, using AI as a tool similar to a calculator.
Compared to using AI to generate commercials or music videos, Jim still prefers using AI to tell stories. For him, “narrative is the original intention of film creation.”
A few years ago, when Jim was self-learning scriptwriting, he not only read books but also practiced by writing scripts and completing them, “If you don’t finish, it doesn’t mean much.”
At the same time, he watched many movies, constantly watching new ones, organizing interesting shots, and watching others’ explanations and analyses. He says he “doesn’t have a solid foundation,” but he is sensitive to images, good at connecting similar objects, and better at using camera language to accurately convey emotions and feelings.
Now, Jim is taking a break, planning to share AI creations, brewing the next AI project, and catching up on the latest developments in each AI video tool. In his view, no matter how many AI tools there are, the resources each person can invest are equal and indispensable. Quantity doesn’t matter; finding the right path for oneself is more important.
In Jim’s Bilibili comment section, a viewer described their viewing experience as “soulless tools, soulful creation.” He replied, “This statement carries a lot of weight.”
The best way to combat anxiety is creation. Jim doesn’t want to speculate on what AI can do in the future or who it will replace. He prefers to believe that constantly creating new works in his hands is a person’s solid fortress.
Source from ifanr
Disclaimer: The information set forth above is provided by ifanr.com, independently of Alibaba.com. Alibaba.com makes no representation and warranties as to the quality and reliability of the seller and products. Alibaba.com expressly disclaims any liability for breaches pertaining to the copyright of content.
