The Human Touch vs. AI: A Conceptual Framework for Character-Based Emotional Engagement in Hybrid Cinema

Hybrid cinema, defined by the integration of live-action footage with animated elements, has evolved into a sophisticated mode of visual storytelling. Traditionally, these works have relied on 2D hand-drawn animation, a medium celebrated for its expressive stylization and what many call “intentional imperfection.” As argued in [1], hand-drawn images often carry the “presence of the animator,” where each line serves as a direct index of human movement and creative choice. This aligns with the perspective in [2] on the “impression of reality,” where the movement in cinema—regardless of its source—serves as a powerful index that anchors the viewer’s belief in the unfolding world. These hand-drawn traces serve as a visual bridge between physical reality and the artist’s imagination. In contrast, the rapid ascent of generative Artificial Intelligence (AI) has introduced a paradigm shift. Here, visual outputs are synthesized through algorithmic processes within latent space systems rather than direct human gestures. This technological evolution creates a profound conceptual and aesthetic tension. On one side, 2D hand-drawn animation embodies the “Human Touch”—a manifestation of artistic intentionality and the visible labor of human authorship. On the other side, AI-generated animation reflects what might be called the “Ghost in the Machine,” where imagery emerges from computational probabilities. In [3], it is suggested that AI-driven media operates through “probabilistic recombination,” appearing fluid yet inherently detached from traditional human craftmanship. This shift suggests that the visual form is no longer a direct result of human labor but a product of statistical predictions. These contrasting paradigms raise critical questions regarding the spectator’s journey. Visual style is not merely an aesthetic choice; it is a fundamental driver of character-based emotional engagement—the affective bond and empathy formed between the viewer and the protagonist. Moreover, visual style influences narrative comprehension, dictating how clearly an audience can decode plot progression and character motivations within a hybrid space. By proposing a blind screening methodology, this study aims to investigate whether the inherent “human” or “machine” qualities within the visual form itself are sufficient to alter audience interpretation, independent of any prior knowledge regarding the production method. Despite the proliferation of generative AI tools, there is a significant gap in research concerning how AI-generated animation affects the essential experience of storytelling. This study addresses this gap by proposing a conceptual framework that examines the intersection of animation technique and audience meaning-making in hybrid cinema. Despite the growing adoption of generative AI in animation production, limited research has examined how AI-generated visual forms influence audience engagement within hybrid cinematic environments. Existing studies primarily focus on technical capability, production efficiency, or audience perception under “informed” conditions, where viewers already know the production method; far less attention has been paid to how viewers respond when animation technique is evaluated solely through its visual characteristics, with contextual information removed. This study addresses that gap by proposing a conceptual framework that examines the intersection of animation technique and audience meaning-making in hybrid cinema. It contributes to the literature in three ways: first, by proposing a conceptual framework linking animation technique and audience engagement in hybrid cinema; second, by introducing perceived authorship as a mediating construct between visual production method and audience response; and third, by extending discussions of authenticity, indexicality, providing a foundation for future empirical investigation. To address these objectives, the study proposes the following research questions and corresponding hypotheses: • RQ1: How do hand-drawn and AI-generated animation techniques influence character-based emotional engagement in hybrid cinema? • H1: Character-based emotional engagement (emphatic concern, perceived vitality, aesthetic intimacy) will be rated higher for the hand-drawn (“Human Touch”) version than for the AI-generated (“Ghost in the Machine”) version, even under blind screening conditions. • RQ2: To what extent does perceived authorship mediate audience responses toward animated characters? • H2: Perceived authorship will mediate the relationship between animation technique and character-based emotional engagement, such that visual cues interpreted as “human-made” will predict higher emotional engagement independent of the film’s actual production method. • RQ3: How do different animation techniques affect narrative comprehension within hybrid cinematic storytelling? • H3: Narrative comprehension will be lower for the AI-generated version than for the hand-drawn version, driven by perceptual inconsistencies, for example, flickering or unstable character details, characteristic of current diffusion-based generation.

A. Audience Perception, Emotional Value, and Authenticity in the AI Era Recent research on audience response to AI-generated media highlights a persistent tension between technological efficiency and perceived authenticity. It was found in [4] that while audiences may acknowledge the technical skill behind computer-generated art, they often exhibit a bias against it when they believe the creative “soul” or human intentionality is absent. This is further supported by [5], which identified a “perception bias” where the mere label of “AI” can lead viewers to undervalue the emotional depth of work. While [6] argues that algorithmic works lack the inherent human labor that audiences associate with authenticity, the “authenticity gap” remains a critical hurdle for generative tools in high-stakes storytelling. B. The Evolution of the Uncanny Valley and Motion Aesthetics The concept of the Uncanny Valley, originally proposed in [7], remains foundational in understanding the psychological discomfort triggered by near-human replicas. In the context of hybrid cinema, this effect is often linked to the consistency of movement; a review of empirical evidence in [8] suggests that “perceptual mismatch”—such as the jittery or unstable motion often found in early generative video—is a primary trigger for the uncanny sensation. This disrupts what is called “perceptual realism” in [9], where a viewer’s ability to accept a fictional character depends not on its photorealism, but on the organic consistency of its motion and interaction within a visual space. C. Human Touch and Indexicality In the realm of traditional animation, [1] established the “presence of the animator” as a key factor in audience connection, where each line serves as a direct index of human movement. In the digital age, this indexicality is challenged. It is argued in [2] that cinema’s power lies in the “impression of reality” through movement, yet when that movement is synthesized via probabilistic models rather than human gestures, the “aesthetic intimacy” traditionally found in hand-drawn imperfections may be lost. D. Technical Constraints and Narrative Meaning-Making The narrative impact of AI’s technical limitations is evident in current generative models. While industrial benchmarks like Sora [10] demonstrate significant leaps in video synthesis, maintaining “temporal consistency” and character constancy across different shots remains a significant challenge. This is academically reinforced in [11] by the creators of VideoFactory, who argue that achieving “spatiotemporal” consistency is a primary hurdle in diffusion-based models. These “latent space” inconsistencies, often manifesting as flickering or unstable character details, can hinder narrative comprehension if the viewer cannot follow the visual logic or the protagonist’s emotional arc due to shifting visual details.

This study adopts a qualitative comparative approach within a practice-based research (PaR) framework to test the proposed conceptual model. The research is structured into three primary phases: production, screening, and analysis. A. The Core Conceptual Framework Mapping The framework positions animation technique as the independent variable influencing audience experience through its intrinsic visual qualities. In this model, perceived authorship functions as a mediating variable that shapes how viewers interpret the creative origin of the animated content. Character-based emotional engagement serves as the primary dependent variable, reflecting the extent to which audiences establish affective connections with animated characters. Furthermore, narrative comprehension is considered an additional outcome variable, representing the audience’s ability to understand plot progression, character motivation, and narrative coherence. The structural mapping of these relationships is illustrated in Figure 1. B. Film Production Design The production phase involves creating two distinct versions of a hybrid film that share the same foundation. Both versions originate from a shared core: a single narrative script and identical live-action footage. Version A utilizes 2D hand-drawn animation to represent the "Human Touch," where each frame is manually stylized. In contrast, Version B represents the "Ghost in the Machine," using AI-generated animation created through diffusion-based video stylization. To isolate visual technique as the primary variable, Version B is prompted to strictly mimic the aesthetic style of Version A, ensuring that both versions feature the exact same duration, pacing, and narrative structure. C. Screening Strategy: Blind Screening The core of this methodology is a blind screening strategy, where participants will view the films without any prior information regarding the animation techniques used. This approach is essential to capture pure audience responses regarding perceived authorship and emotional engagement. Data collection will be conducted using open-ended questionnaires to capture detailed emotional reactions and Likert-scale ratings to measure the dimensions of audience response, including Empathic Concern, Perceived Vitality, Aesthetic Intimacy, and Narrative Comprehension.

A. The Mechanism of Influence The framework operates on the premise that the independent variable (animation technique) dictates the physical properties of the film. These physical properties are expected to influence the mediating variable, which is perceived authorship. This represents the audience's subconscious or conscious attribution of the creative source—whether the viewer perceives the visuals as having a "human" or "machine" origin based purely on the visual cues. B. Expected Analysis of Character-Based Emotional Engagement The framework proposes that these variables directly impact the dependent variable, projecting distinct evaluative dimensions: • Empathic Concern: The analysis explores the degree of emotional alignment and empathy the audience feels toward the character’s situation. The manual intentionality of Version A is expected to project an indexical trace of human labor that stabilizes empathy, whereas Version B’s statistical fluidness must be examined to see if latent variations disrupt this emotional bond. • Perceived Vitality: This measures the perceived liveliness, soul, and expressive movement of the character within the hybrid environment. The study analyzes whether the organic imperfections of traditional drawing enhance vitality, or if algorithmic motion dynamics can simulate an acceptable form of artificial liveliness. • Aesthetic Intimacy: This dimension analyzes the sense of closeness and personal connection created through the specific texture and visual form of the animation. It addresses whether high-frequency machine artifacts or manual pencil lines generate higher proximity in audience perception. • Narrative Comprehension: This dimension evaluates how effectively audiences understand plot progression, character motivation, and overall narrative coherence. The analysis examines whether differences in animation technique influence viewers’ interpretation of narrative events within the hybrid cinematic environment.

Despite the rapid development of generative AI in animation, its impact on the fundamental aspects of storytelling—particularly audience emotional engagement—remains insufficiently understood. Existing studies often focus on technical performance or audience bias under "informed" conditions, where viewers are already aware of the production method. However, limited research examines how different animation techniques influence audience response when such contextual information is removed. This creates a gap in understanding whether the inherent visual qualities of animation alone are sufficient to shape audience perception. Therefore, by implementing this conceptual framework within a blind screening context, the study evaluates the perceptual impact of visual form independent of external knowledge bias. The theoretical bridge established between computer-generated visuals and classic film theories—such as indexicality in [2] and the Uncanny Valley in [7]—helps redefine the limits of technological automation. It allows filmmakers to understand that visual style is not merely an aesthetic choice, but a critical tool that dictates how clearly an audience can decode plot progression, attribute authorship, and establish deep psychological connections with a protagonist within a hybrid space. As a conceptual framework, however, this study’s propositions remain theoretical pending empirical validation through the planned production and screening phases. Several limitations should be acknowledged for future work. First, the practice-based design relies on a single narrative and a small planned sample of blind-screening participants, which may limit generalizability across genres and cultural contexts. Second, matching Version B’s AI-generated aesthetic strictly to Version A's hand-drawn style may itself introduce a confound, as current diffusion-based tools vary in their ability to faithfully mimic a target style, potentially conflating “AI origin” effects with “stylization fidelity” effects. Third, self-reported Likert and open-ended measures of empathy and vitality are inherently subjective and may be influenced by participants’ pre-existing attitudes toward AI, which the blind design mitigates but can not fully eliminate. Finally, rapid advancement in generative video models means findings based on current AI capabilities may have a limited shelf life as temporal and spatial consistency continue to improve.

This study contributes to academic and practical discussions by proposing a conceptual framework that isolates visual form as a primary driver of audience response. It provides a refined analytical model connecting visual style, perceived authorship, and emotional meaning-making. Furthermore, it offers practical implications for filmmakers navigating the integration of AI while trying to maintain the emotional "heart" of their storytelling. In conclusion, this paper has presented a conceptual framework for analyzing audience response in hybrid cinema. By focusing on the contrast between the "Human Touch" and the "Ghost in the Machine," the study emphasizes the role of intrinsic visual qualities in shaping character-based emotional engagement and narrative comprehension within hybrid cinematic environments. Future work will involve the actual production of the two film versions to generate empirical findings. As generative AI continues to redefine storytelling, understanding how audiences respond to its aesthetic qualities—independent of prior knowledge—will become increasingly important for the future of the medium.