INTRODUCTION
The film industry has experienced multiple transformative phases, including the transition from silent to sound cinema, from traditional film shooting to digital cinematography, from exclusive theatrical exhibition to multi- platform media screening, and most recently, from conventional intelligent film production to AI-driven intelligent filmmaking. Each of these revolutions has followed a similar trajectory: beginning with technical challenges,
788 Chen progressing through the integration and co-creation of art and technology, and culminating in the redefinition of artistic concepts and aesthetic paradigms. The AI intelligent production of short films emerges at a pivotal historical moment marked by media transformation and represents a significant phase in the evolution of Chinese cinema. Chinese media professionals have not only identified the cutting-edge trends in AI-driven film production but have also recognized both the creative potential and the disruptive impact of AI-generated content. AI has effectively dismantled traditional technical and financial barriers, while challenging long-held perceptions about authorship, creativity, and artistic value. Looking ahead, future art creators must acknowledge that the convergence of art and technology represents the trajectory of artistic innovation. Artists are thus encouraged to embrace artificial intelligence not merely as a tool, but as a collaborative partner in the creative process, redefining the boundaries of cinematic expression.
The rationale for selecting this research topic—namely, the exploration of the migration of author-directors’ visual aesthetic styles in AI-generated short film creation—stems from the observation that contemporary renowned auteurs are engaged in a dual negotiation process in their creat. On one hand, they strive to reconcile the cinematic traditions inherited from their predecessors with their own distinctive modes of artistic expression. On the other hand, they are confronted with the challenge of integrating their hands-on filmmaking experience with the algorithmic logic underlying AI-generated imagery, as well as navigating the interplay between machine- generated outputs and the spontaneity of human creative inspiration. This paper highlights the creative and ethical dilemmas arising from the emergence and potential erosion of individuality and uniqueness in AI-generated content. Rather than beginning with the technical mechanisms of AI image generation as quasi-entities, this research focuses on AI-generated short films created by directors known for their distinct authorial styles. Based on the technical logic of keyword-driven creation, it explores the evolution of directors’ narrative approaches, the aesthetic framework of AI-generated visual art, and the broader implications for artistic reflection. It investigates the transformation of aesthetic standards, the challenges of intelligent visual representation, and the expanded expressive potential of contemporary image-making practices.
THE CURRENT APPLICATION STATUS AND DEVELOPMENT PREDICAMENTS OF AUTEURS’ AESTHETIC STYLE TRANSFER IN FILM CREATION BY AI At the end of 2024, Kuaishou launched the “China’s First AIGC Director Co-Creation Program,” marking a significant step in the integration of artificial intelligence and cinematic art. Through its self-developed video generation large model, “Keling AI,” Kuaishou collaborated with nine acclaimed directors—Li Shaohong, Jia Zhangke, Ye Jintian, Xue Xiaolu, Yu Baimei, Dong Runnian, Zhang Chiyu, Wang Zichuan, and Wang Maomao (Tuziksi)—to explore the potential of large AI models in film-level content creation. This initiative represents a pioneering effort to merge the artistic styles of auteur directors with AI-driven filmmaking technologies. The most
Exploratory Research on the Migration of Auteurs’ Cinematic Aesthetic Styles 789 notable innovation lies in the unregulated interaction and experimental fusion between directorial artistic expression and AI-generated imagery, as well as the exploration of how intelligent technologies can expand the expressive boundaries of visual art. However, one of the primary challenges encountered during the production process was the inconsistency and unpredictability of AI-generated content. The films utilized AI’s capabilities in simulating physical environments and generating complex visual concepts, relying on Keling AI to produce the final video sequences. The result was nine short films, each approximately three minutes in length, which collectively demonstrate the evolving migration of auteurs’ cinematic aesthetic styles within the context of AI-generated filmmaking.
The Current Application Status of Auteurs’ Cinematic Aesthetic Styles in AI-Generated Short Films AI-generated short films production encompasses three advanced modes: AI-assisted production, AI-assisted creation, and AI autonomous creation. The short films analyzed in this study exemplify a collaborative model in which the auteurs’ creative vision are integrated with AI-assisted creation. Autonomous generation, automated production workflows, and self- optimizing learning mechanisms collectively define the core characteristics of AI-generated filmmaking. The ability to migrate auteurs’ cinematic aesthetic styles represents a critical capability that AI-generated production must further develop. From the perspective of aesthetic innovation in AI-generated film content, AI systems can be trained using an auteur’s body of previous work to internalize and replicate their visual language. During the scriptwriting phase, AI can generate screenplays and character dialogues that align with the auteur’s distinctive visual and narrative style. In the pre-production and visual concept development stages, AI can produce art direction scenes, concept sketches, storyboards, and shot lists that reflect the auteur’s aesthetic sensibilities. In the production phase, AI facilitates the generation of 3D digital assets consistent with the auteur’s visual style, enables the construction and interaction with virtual environments, and supports the creation of virtual characters, actions, dialogues, and sound effects tailored to the director’s artistic intent. In post-production, AI can perform intelligent chroma keying, color grading, and noise reduction on video clips that embody the auteur’s aesthetic. It can also restore and enhance damaged footage, simulate and reconstruct film music and sound effects, and apply post-production stylization transfer techniques to implement visual effects that align with the auteurs’ cinematic identity. By integrating multiple AI-driven methodologies across pre-production, production, and post-production workflows, it becomes feasible to effectively transfer and preserve the aesthetic style of individual auteurs within AI-generated short films.
AI-generated short films are currently undergoing rapid technological and artistic advancements. In the domain of image generation, notable progress has been made in the realism of characters’ facial expressions, body movements, and voice modulation. The instability associated with randomly
790 Chen generated imagery has been significantly reduced, and the controllability of video generation has improved to a considerable extent. However, the effectiveness of transferring auteurs’ cinematic aesthetic styles remains an area requiring further refinement. Moreover, the integration of multi-modal video generation elements—including text, images, videos, and 3D assets— has demonstrated increasing sophistication. From an artistic perspective, the rendering of fluid elements such as river water, and ocean waves has become more realistic. The simulation of natural and artificial lighting effects has also become more vivid and nuanced. The depiction of fine details such as human hair and animal fur has reached a higher level of realism. The creation of atmospheric effects using various materials has become more diverse and expressive. Additionally, scene composition, camera movement, and transitions between scenes have become more dynamic and visually complex. Despite these advancements, the methodologies for transferring auteurs’ cinematic aesthetic styles remain in the exploratory and developmental stages, requiring further theoretical and technical refinement to achieve mature implementation.
The Development Predicaments of Auteurs’ Cinematic Aesthetic Styles in AI-Generated Short Films Currently, there are notable limitations and challenges in the application of AI-generated short films to the cinematic aesthetic styles of auteurs within the domain of film production. These constraints are primarily manifested in the quality indicators related to video and audio generation. The essence of cinema lies in its distinctive cinematic quality, which encompasses a range of visual and auditory characteristics. From a visual perspective, cinematic imagery typically features high brightness, resolutions of 4K or higher, frame rates of 60 or 120 frames per second, wide color gamuts, and high dynamic ranges—collectively enabling the creation of visually complex and immersive 3D imagery. From an auditory standpoint, cinema offers a spectrum of sound experiences, ranging from conventional surround sound and Dolby audio effects to advanced immersive audio technologies. These elements contribute to a heightened sense of realism and immersion in both visual and auditory dimensions. Furthermore, specialized formats such as IMAX, CINITY, 4D films, and 360-degree panoramic films extend the cinematic experience by integrating visual, auditory, tactile, and even olfactory stimuli. This multisensory engagement enhances audience immersion and emotional connection to the content. The adaptation and migration of auteurs’ cinematic aesthetic styles can also be evaluated and manifested through these technical specifications and sensory frameworks.
In contrast, AI-generated short films currently face significant technical and artistic limitations that hinder their application in professional cinematic contexts. First, the generation of such films necessitates substantial computational resources and processing power. At present, the output duration of AI-generated content remains confined to mere seconds, falling far short of the standard 90-minute runtime required for theatrical feature films. Second, the technical specifications of AI-generated imagery—such as resolution, frame rate, color gamut, and dynamic range—fail to meet the
Exploratory Research on the Migration of Auteurs’ Cinematic Aesthetic Styles 791 industry standards for cinematic release. As a result, these images require extensive post-processing to be integrated into existing workflows, and even then, their compatibility with high-format cinematic systems remains limited. Third, AI-generated films often struggle to accurately represent complex human actions and physical interactions that conform to real-world logic and spatial dynamics. This highlights the need for significant improvements in spatial recognition and motion modeling capabilities. Fourth, as an art form rooted in human experience, cinema relies on the nuanced portrayal of characters to convey auteurs’ cinematic aesthetic styles. Therefore, the authenticity of characters’ expressions, actions, and emotional depth constitutes a critical criterion for evaluating film quality. Current AI-generated characters lack the level of realism required for theatrical presentation, thereby undermining the effective transmission of auteur-driven aesthetics. Furthermore, while AI can generate narratives with intricate plot structures, it remains incapable of embedding the complex emotional depth necessary to portray multifaceted human nature. Consequently, AI-generated films fall short in achieving profound storytelling and eliciting meaningful emotional engagement from audiences.
JIA ZHANGKE’S EXPLORATION OF AI-GENERATED SHORT FILMS Director Jia Zhangke centers his cinematic exploration on the lived experiences of ordinary individuals, with particular emphasis on the socio-cultural realities of marginalized groups—such as disillusioned youth in provincial towns and those who find themselves dislocated by the rapid transformations of modern Chinese society. His preference for a documentary-inspired audiovisual style not only captures the authenticity of everyday life but also stimulates critical discourse regarding the relationship between camera presence and perceived objectivity. Collaborating closely with cinematographer Yu Liwei, a long-standing creative partner, Jia employs a visual language rooted in traditional filmmaking techniques, yet adapted to reflect the evolving urban landscapes of contemporary China. This approach establishes a distinctive aesthetic grounded in cinematic realism. Moreover, Jia frequently integrates archival media and historical audiovisual materials that encapsulate the zeitgeist of specific historical periods. These elements, alongside his portrayal of individuals dancing in discotheques, serve as visual metaphors for the emotional and psychological states of a generation. Occasionally, surrealist motifs emerge within his work, offering a creative counterpoint that reflects the director’s artistic sensibilities and conceptual intentions. In the domain of sound design, Jia demonstrates expertise in on-location audio capture, incorporating authentic ambient sound and voice-over narration. He also strategically embeds popular music that aligns with the emotional arc of the narrative, thereby enhancing the film’s affective resonance.
Jia Zhangke’s first AI-generated short film, *Harvesting Wheat*, continues the thematic exploration established in his *Hometown Trilogy*. It reflects the idea that individuals often come to truly understand the concept of “hometown” and experience nostalgia only after physically and emotionally distancing themselves from it. The film retains Jia’s long-standing focus on the lived experiences of ordinary Chinese people within the broader context
792 Chen of societal and technological transformation. It further examines universal themes present in his body of work, including familial bonds, existential identity, belonging, personal value, life aspirations, and the search for meaning. *Harvesting Wheat* employs a road movie narrative structure, centered around a robotic protagonist who connects various characters throughout the film. Through fragmented vignettes, the film illustrates the evolving relationships between humans and silicon-based entities, offering a contemporary reflection on how ordinary individuals navigate communication and coexistence with artificial beings in the technologically saturated landscape of the 2020s. As the director himself has explained, the short film aims to investigate two key aspects of AI-generated filmmaking. First, it tests the AI’s capacity to reconstruct realistic environments, with particular attention to the authenticity and consistency of spatial and visual representation. The ability to create a convincing sense of place remains a crucial indicator of a director’s visual storytelling proficiency. Second, the film evaluates AI’s capability to maintain the coherence and consistency of a central character throughout different scenes. The silicon-based character was generated by training the AI model on a dataset comprising numerous photographs and extensive video footage of synthetic human-like figures. The creators issued specific instructions to Keling AI to generate a male character with predefined physical features. They then produced a series of test clips to assess the continuity and visual coherence of the character when placed in varying narrative contexts, thereby exploring techniques for preserving character consistency across diverse scenes.
In addition, Jia Zhangke has stated that the purpose of this work is to explore the current potential of AI in autonomous image creation and to demonstrate the challenges and creative boundaries that AI still faces. The road movie genre and its visual style provide an effective framework for examining the behavioral and emotional states of characters across diverse application scenarios. In *Harvesting Wheat*, AI-generated main characters remain present throughout the narrative, moving through various spatial settings such as coffee shops, urban sidewalks, intercity highways, and wheat fields. These characters engage in verbal and behavioral interactions with others, including the task issuer, a robot walking with a child, an autonomous vehicle, an elderly person seeking roadside assistance and their robotic attendant, as well as the parents of the task issuer. Maintaining a consistent and coherent main character was the most significant challenge in the short film. The production made a concerted effort to ensure continuity of the protagonist across different contexts, successfully sustaining character presence throughout the six-minute runtime. The rendering of character details—such as skin texture, hair, and lighting effects—highlighted the subjective visual quality achievable through AI-generated short film technology. Jia Zhangke is known for his use of long takes and wide or medium-wide shots to capture movement and convey the passage of time through a dynamically framed visual narrative. However, due to current technical limitations in AI-generated filmmaking, *Harvesting Wheat* employed traditional cinematographic techniques in each scene, including shot-reverse-shot sequences, spatial blocking, and actor choreography to
Exploratory Research on the Migration of Auteurs’ Cinematic Aesthetic Styles 793 realize the director’s creative vision. Furthermore, the use of slow-motion, as well as pan, tilt, zoom, and tracking shots, contributed to a heightened cinematic sensibility. In terms of spatial representation, Keling AI successfully established a coherent sense of spatial orientation. Notably, the film did not rely on special effects or post-production techniques typical of traditional filmmaking; all visual elements were generated entirely through AI.
LI SHAOHONG’S EXPLORATION OF AI-GENERATED SHORT FILMS Films directed by female directors that centers on the authentic experiences and social conditions of women can be categorized into three principal types: films with female themes or protagonists, feminist-oriented films, and films directed by women. Li Shaohong, recognized as one of the most prominent female directors of the Fifth Generation, has progressively expanded her creative boundaries throughout her career. Beginning with her early avant-garde visual style, she transitioned to thematic explorations of female consciousness and narrative perspectives during her artistic evolution, and most recently, she has ventured into AI-generated short films—creating character-driven, plotless vignettes. In Li’s filmography, the interplay between dreams and reality serves as a narrative device to externalize the identity crises and emotional struggles faced by female characters. The female figures she constructs challenge the conventional stereotypes of self-sacrifice and passive benevolence. Through their experiences of navigating the tensions between societal expectations and personal aspirations, these characters develop into resilient, independent, and self-aware individuals, embodying a new paradigm of female subjectivity in contemporary Chinese cinema.
Li Shaohong’s first AI experimental short film, *Flowers Fill the Shoal*, created with Keling AI, is a narrative without a conventional plot, characterized by the absence of a strict causal chain or dramatic structure. The film represents an exploration of temporality—how time is perceived, constructed, and represented—and serves as a bridge connecting Li’s earlier cinematic works. It functions as a test of AI’s creative potential as an emerging technological tool in filmmaking. The title of the film is inspired by the classical Chinese poem *Fisherfolk – A Rowing Boat in the Spring Breeze and a Leafy Boat* by Li Yu, the last ruler of the Southern Tang Dynasty. The poem depicts a fisherman enjoying the spring breeze while rowing a boat, fishing, and drinking wine. The islet is blanketed with blooming flowers, and the wine cup is full in hand. Immersed in the vast expanse of waves, the fisherman experiences a profound sense of freedom. The protagonist of the short film is a female director, and the narrative unfolds through her subjective perspective. She embodies the image of a modern woman— assertive, ambitious, and committed to prioritizing her personal life planning. As a female director living a female experience, she reflects on and internalizes the tragic destinies of women across different historical periods under patriarchal structures. The voice-over features the female director stating: “Before becoming a director and telling their stories well, one must first become them.” This reflects the journey women have undergone—from a state of suppressed female consciousness, through awareness, to conscious self-recognition, and ultimately toward self-liberation.
794 Chen The female director engages in a cross-temporal dialogue with female characters from Li Shaohong’s earlier films through a voiceover, symbolizing a journey across historical and emotional landscapes. Visual motifs of freedom—such as kites without strings—glide through the spatial domains of these female protagonists, metaphorically illustrating that love is as beautiful as it is fragile, much like a kite soaring in the sky. This imagery reflects women’s enduring struggle to break free from the oppressive grip of feudal patriarchy. Through her cinematic narratives of love, Li conveys a critical message: love cannot liberate women, nor can it serve as a means of self- redemption. Women are portrayed as birds ensnared in an iron net constructed from seemingly positive concepts—survival, family, power, and love. The female protagonists sequentially wander and run through the filmic spaces drawn from Li’s earlier works, embodying the emotional and existential anxieties of women across generations. As Li explained in an interview, she has expanded the boundaries of visual imagination and narrative depth by integrating AI into the construction of cinematic storytelling and artistic vision. In terms of technical execution, the short film strategically leverages the strengths of AI while mitigating its limitations. Given the inherent instability of AI-generated imagery, the film employs multiple spatial settings to construct its narrative. Each space corresponds to a distinct historical era and reflects the evolution of female characters throughout Li’s directorial career. By interweaving narratives across different times through character- driven storytelling, the film conveys a unifying theme: the fate of women is perpetually entangled with the dual pursuit of self-identity and freedom. Despite being constrained by survival pressures, social conditioning, objectification, definition by others, and confinement by love, women’s fates continue to repeat across historical cycles. Yet, generation after generation, women persist in questioning and exploring, driven by the belief that life is a vast wilderness—not a predetermined track.
Li Shaohong excels at externalizing the complex inner worlds of female characters—marked by intense desire and vibrant vitality—through eerie, spectral, and even surreal imagery. These visual strategies, embedded within narrative-driven audiovisual elements, leave a profound and lasting impression on the audience. The AI experimental short film *Flowers Fill the Marsh* exemplifies Li’s avant-garde visual style and reflects the creative consciousness of a female auteur. The film’s high degree of stylistic coherence, visual stability, and fidelity to the original prompt are particularly striking. A critical question in the evolving convergence of AI technology and cinematic art is whether AI can effectively emulate the visual aesthetics and directorial techniques of individual filmmakers, and achieve a meaningful transfer of their unique cinematic language. By leveraging large language models and analyzing detailed parameters from existing films—including focal length, aperture, shutter speed, shot duration, and camera movement trajectories— AI can begin to reconstruct complex visual elements such as depth of field, color grading, and motion blur. This enables the replication of original visual styles within AI-generated virtual environments, opening new possibilities for stylistic continuity and innovation in digital filmmaking.
Exploratory Research on the Migration of Auteurs’ Cinematic Aesthetic Styles 795 YU BAI MEI’S EXPLORATION OF AI-GENERATED SHORT FILMS
Unlike many female directors whose works often emphasize emotional depth and social critique, Yu Baimei is primarily recognized for his comedic expertise. This section will examine how AI technology can be utilized to enhance comedic effects, drawing on the creative practices of Yu Baimei as representative example.
In the AI short film, the director integrates his signature comedic elements with the portrayal of Zhong Kui, a well-known mythological figure, to highlight the protagonist’s chivalric spirit. The AI-generated visuals effectively construct a suspenseful and mysterious atmosphere. Following the opening credits, the moon appears first, followed by a camera pan to cold, gleaming weapons, and finally settling on the protagonist’s back. The film alternates between close-ups of body parts and full back views, enhancing the visual rhythm and tension. The final scene depicts Zhong Kui clad in a red battle robe, wielding sword and knife, moving through a dark, mist-laden mountain forest. Subsequently, three demons—the rabbit demon, the toad demon, and the tree demon—appear one after another, each swiftly defeated by Zhong. The film uses the interactions between Zhong and these demons to create dramatic contrasts. Initially, Zhong menacingly points his sword at a seemingly innocent rabbit, only to later lower his guard and gently lift it. The rabbit then sheds its cute disguise, revealing a fierce gaze before transforming into a giant rabbit demon. Zhong’s ability to detect this deception illustrates his wisdom. The toad demon approaches with a seemingly friendly demeanor but suddenly attacks Zhong using gold coins as weapons. Zhong defeats the toad with his sword. The tree demon then conjures a barrier, within which the visuals shift to a black-and-white ink painting style. Zhong first engages in hand-to-hand combat with owl demons before summoning his signature weapon to destroy the barrier. As he strikes, Zhong declares: “The heavens have lost their color. How dare you trap me in this barrier? See my sword— no ghosts shall disturb. Break!” His ability to overcome the tree demon demonstrates his courage. In a climactic moment, Zhong leaps toward the camera, delivering a powerful strike. The camera then pans to the moon in the background, which morphs into a smartphone screen displaying the “Do Not Disturb” mode, symbolizing a holiday message: “No ghosts shall disturb.” This AI-generated visual metaphor effectively bridges mythological storytelling with contemporary life. The director aims to use traditional cinematic techniques while employing AI as a tool to explore the boundaries of live-action imaging technology. By applying text-to-image generation and first-and-last frame imaging methods, the film establishes a connection between AI and physical performance. Yu Baimei also intends for this short film experiment to investigate the potential of AI in conveying artistic and emotional depth.
The AI-generated short film demonstrates a high degree of overall visual and narrative coherence. As the central protagonist, Zhong Kui maintains consistent facial modeling and character design throughout the film. The setting—a deserted forest under moonlight—effectively reinforces the atmospheric consistency and enhances the visual tone. While the narrative
796 Chen primarily centers on Zhong’s battle against supernatural entities, the character’s physical movements are deliberately restrained in amplitude, contributing to a sense of smooth and controlled motion. Additionally, the strategic use of push-in and pull-out shots functions as an external cinematic force that amplifies the visual impact and spatial dynamics of the character’s actions. The film exhibits strong aesthetic and technical imaging quality, successfully realizing the director’s intended color palette and spatial composition. Its visual style is deeply rooted in traditional Chinese mythological aesthetics, reflecting a distinctive cultural visual identity. This AI-assisted production not only achieves technical coherence but also preserves a strong artistic vision that bridges classical iconography with contemporary digital filmmaking techniques.
CONCLUSION
This study investigates how a director’s visual aesthetics can be effectively integrated into AI-generated film production while adhering to legal regulations, ethical guidelines, technical standards, and communication efficiency. Building upon the artistic characteristics of the director’s prior works, this paper analyzes the presentation and reception of AI-generated short films, as well as the methods and outcomes of translating traditional directorial aesthetics into an AI-driven context. Given that AI short film creation should be categorized according to genre conventions, these works aim to engage younger audiences by delivering emotional and experiential value. Furthermore, a balanced cognitive framework is necessary to bridge the gap between AI-assisted authorship and audience reception. The experimental nature of AI short films enables creators to gather valuable market feedback and better understand evolving viewer preferences. AI short films are increasingly drawing inspiration from short video production, which benefits from a rapid production cycle and high responsiveness to current audience interests. Additionally, AI short films are actively exploring strategies to enhance the coordination among key cinematic elements—such as sound design, art direction, and visual effects—to achieve greater creative efficiency and technical integration.
ACKNOWLEDGMENT
The authors would like to acknowledge the support of the 2018 Beijing Social Science Foundation Research Base General Project “Research on the National Image Creation of the Chinese Film School in the Digital Film Age” (Project Number: 18JDYTB004).