Abstract
The integration of artificial intelligence (AI) and Virtual Production (VP) into filmmaking marks a transformative moment in media production. These technologies are not merely tools for cost-efficiency or labor reduction; they are reshaping the very ontology of filmmaking—particularly its collaborative nature. Current AI-VP production is highly experimental, marked by a “try-and-hope-it-works” phase of technological trial-and-error. It is within this shifting terrain that this paper argues for the “Imagined Audience” as a crucial interdisciplinary conceptual lens for understanding how film labor, authorship, aesthetics, and institutional hierarchies are being redefined.
In this new environment, the imagined audience functions as both a ghostly presence and a structuring force—emerging, disappearing, and reemerging during every stage of production. It mediates between human and machine, past and future, authorial intention and algorithmic prediction. The imagined audience serves as a theoretical tool to synthesize traditionally fragmented fields: aesthetics, institutional analysis, audience research, labor studies, and technology studies.
This conceptualization draws on foundational work by Gans (1957), who emphasized how creators imagine their audience when making decisions. Ian Ang (1991) argued that audiences are constructed through institutional logics rather than empirically knowable. Litt (2012) extended this to digital environments, where imagined audiences are constantly negotiated. Turow and Draper (2014) detail how audience imaginaries are actively constructed by industries through datafication and segmentation. Taken together, these perspectives lay the foundation for understanding how AI-enhanced filmmaking integrates audience imaginaries into algorithmic design and creative decision-making.
This paper presents a production ethnography of three short sci-fi films created using VP and AI. One was directed remotely from Suzhou, China, with a VP studio in Seoul; the other two involved co-production between teams in Singapore and Korea. In all cases, real-time manipulation of virtual environments using ICVFX, AI-assisted design tools, and game engines enabled new forms of collaboration among directors, VADs, VFX supervisors, game designers, and AI developers. These workflows collapsed traditional temporal distinctions between pre-production, shooting, and post-production, creating a continuous feedback loop shaped by both human vision and machine prediction.
These changes destabilized established hierarchies within film crews. For instance, VAD teams assumed roles traditionally held by cinematographers, while directors modified scenes live based on algorithmically inferred viewer retention curves. Women accessed cinematographic and design roles previously out of reach, highlighting VP’s potential to disrupt entrenched gender hierarchies in media labor. In such a setting, creative machines do not merely assist; they participate in the negotiation of meaning. This creates a tug-of-war between the audience imagined by human creators and the audience imagined by AI—a dynamic that becomes central to the production process itself. In fact, generative tools like image generators and language models engage in their own form of audience prediction, not for the public, but for the human creator as the immediate user. In these iterative loops, the machine's imagined audience is the creator, and the goal is to approximate the creator’s expectations through each new version. This almost endless back-and-forth—try, prompt, adjust, regenerate—manifests as a form of ghostly dialogue between human and machine—where the imagined audience floats, recedes, and reappears throughout the creative process. It is simultaneously a projection, a collaborator, and a feedback mechanism, revealing the complex affective and epistemological entanglements at the heart of AI-mediated production.
Drawing on Williams’ (1974) sociotechnical framework, which positions television as a cultural form shaped by social and institutional forces, and Manovich’s (2001) new media theory, which analyzes how digital technologies alter media logics, I interrogate what qualifies as “new” in contemporary filmmaking. These frameworks help illuminate how AI and VP technologies are not merely additive tools but fundamentally restructure storytelling strategies, aesthetic values, and the hierarchy of creative roles within production ecosystems. These theoretical insights lay the groundwork for examining the implications of AI and VP not only at the level of aesthetics and narrative form but also in how labor is restructured and redistributed across global production workflows.
Building on these shifts, the labor dimension is equally crucial. Roles like script translators, localization experts, and UX designers become central intermediaries between human creators, machines, and inferred audiences. Following Soto-Sanfiel and Villegas-Simón (2024), who show how creators navigate LGBTQ+ representation under institutional pressure, this paper argues that AI-VP production also entails navigating multiple, sometimes contradictory, audience imaginaries—often with ethical consequences.
Caldwell’s (1995) concept of televisuality highlights how television evolved into a medium characterized by overt stylization, production excess, and textual density. Though developed in the context of 1980s American television, this idea helps us think about the aesthetic implications of AI-enhanced cinema. Rather than reinforcing a narrative of seamless efficiency, VP often results in an overload of information and design—aesthetic layers tuned by algorithms, real-time rendering, and predictive analytics. This paper uses Caldwell’s theory not to draw a direct parallel but to foreground how visual self-awareness and layered mediation persist, and even intensify, in AI-driven production contexts. In this way, Caldwell helps frame how imagined audiences are addressed not only through content but through the formal complexity designed to catch and retain fragmented, multitasking viewers.
In this light, the imagined audience framework makes visible the underlying power struggles—between auteurism and collective authorship, between intuition and prediction, between cultural specificity and algorithmic generalization. More importantly, the imagined audience operates as a dynamic mediator among three intersecting powers central to cultural production: the power of economic calculation, the power of aesthetic suggestion, and the power of interpretation. At the industrial level, imagined audiences are shaped by programming strategies, platform metrics, and marketing models that aim to quantify viewer behavior—reflecting the power of economic calculation. At the textual level, they are anticipated through production design, genre conventions, and narrative techniques that seek to elicit emotional resonance—demonstrating the power of aesthetic suggestion. And at the audience level, real viewers interpret content in idiosyncratic and socially contextualized ways—exercising the power of interpretation. By mediating among these forces, the imagined audience becomes a crucial site where institutional goals, creative intent, and interpretive agency converge and contend, making it a nerve center of meaning-making within AI-enhanced media production. It enables a critical understanding of how AI reconfigures authority, creativity, and meaning in contemporary media production.
In conclusion, the imagined audience offers a robust theoretical and methodological tool to analyze the transformation of creative labor and media aesthetics under AI. It highlights how AI is not just a tool but a participant in meaning-making, fundamentally entangled with how creators anticipate, negotiate, and respond to the perceived desires of others—whether human or machine. This paper contributes to rethinking media production as a space where human labor, technological systems, and aesthetic forms converge through shared, contested visions of an audience that may never be fully known but always shapes the story.
References
Andrejevic, M. (2008). Watching television without pity: The productivity of online fans. Television & New Media, 9(1), 24–46. https://doi.org/10.1177/1527476407307241
Ang, I. (1991). Desperately seeking the audience. Routledge.
Bodini, A., Manohar, A., Colecchia, F., Harrison, D., & Garaj, V. (2024). Envisioning the future of virtual production in filmmaking: A remote co-design study. Multimedia Tools and Applications, 83, 19015–19039.
Caldwell, J. T. (1995). Televisuality: Style, crisis, and authority in American television. Rutgers University Press.
Celis Bueno, C., Chow, P. S., & Popowicz, A. (2024). Not "what", but "where is creativity?": Towards a relational-materialist approach to generative AI. AI & Society. https://doi.org/10.1007/s00146-024-01921-3
Ferrari, F., van Dijck, J., & van den Bosch, A. (2023). Observe, inspect, modify: Three conditions for generative AI governance. New Media & Society. https://doi.org/10.1177/14614448231214811
Gans, H. J. (1957). The creator-audience relationship in the mass media. In B. Rosenberg & D. M. White (Eds.), Mass culture: The popular arts in America (pp. 315–321). Free Press.
Guzman, A. L., & Lewis, S. C. (2024). What generative AI means for media industries. Emerging Media. https://doi.org/10.1177/27523543241289239
Lin, B. (2024). The AI chatbot always flirts with me, should I flirt back? From the McDonaldization of friendship to the robotization of love. Social Media + Society, 10(4). https://doi.org/10.1177/20563051241296229
Litt, E. (2012). Knock, knock. Who’s there? The imagined audience. Journal of Broadcasting & Electronic Media, 56(3), 330–345. https://doi.org/10.1080/08838151.2012.705195
Livingstone, T. (2024). Game engines: Optimising VFX, reshaping visual media. NECSUS European Journal of Media Studies, Spring 2024.
Lobato, R., & Lotz, A. D. (2020). Imagining global video: Netflix. Journal of Cinema and Media Studies, 59(3), 132–136. https://doi.org/10.1353/cj.2020.0025
Manovich, L. (2001). The language of new media. MIT Press.
Soto-Sanfiel, M. T., & Villegas-Simón, I. (2024). Scriptwriters’ conceptions of audience attitudes toward LGBTQ+ characters. Mass Communication and Society, 27(5), 1252–1276. https://doi.org/10.1080/15205436.2023.2292123
Turow, J. (2011). The daily you: How the new advertising industry is defining your identity and your worth. Yale University Press.
Turow, J., & Draper, N. (2014). Industry conceptions of audience in the digital space: A research agenda. Cultural Studies, 28(4), 643–656. https://doi.org/10.1080/09502386.2014.888929
van Dijck, J. (2009). Users like you? Theorizing agency in user-generated content. Media, Culture & Society, 31(1), 41–58. https://doi.org/10.1177/0163443708098245
Williams, R. (1974). Television: Technology and cultural form. Schocken Books.
In this new environment, the imagined audience functions as both a ghostly presence and a structuring force—emerging, disappearing, and reemerging during every stage of production. It mediates between human and machine, past and future, authorial intention and algorithmic prediction. The imagined audience serves as a theoretical tool to synthesize traditionally fragmented fields: aesthetics, institutional analysis, audience research, labor studies, and technology studies.
This conceptualization draws on foundational work by Gans (1957), who emphasized how creators imagine their audience when making decisions. Ian Ang (1991) argued that audiences are constructed through institutional logics rather than empirically knowable. Litt (2012) extended this to digital environments, where imagined audiences are constantly negotiated. Turow and Draper (2014) detail how audience imaginaries are actively constructed by industries through datafication and segmentation. Taken together, these perspectives lay the foundation for understanding how AI-enhanced filmmaking integrates audience imaginaries into algorithmic design and creative decision-making.
This paper presents a production ethnography of three short sci-fi films created using VP and AI. One was directed remotely from Suzhou, China, with a VP studio in Seoul; the other two involved co-production between teams in Singapore and Korea. In all cases, real-time manipulation of virtual environments using ICVFX, AI-assisted design tools, and game engines enabled new forms of collaboration among directors, VADs, VFX supervisors, game designers, and AI developers. These workflows collapsed traditional temporal distinctions between pre-production, shooting, and post-production, creating a continuous feedback loop shaped by both human vision and machine prediction.
These changes destabilized established hierarchies within film crews. For instance, VAD teams assumed roles traditionally held by cinematographers, while directors modified scenes live based on algorithmically inferred viewer retention curves. Women accessed cinematographic and design roles previously out of reach, highlighting VP’s potential to disrupt entrenched gender hierarchies in media labor. In such a setting, creative machines do not merely assist; they participate in the negotiation of meaning. This creates a tug-of-war between the audience imagined by human creators and the audience imagined by AI—a dynamic that becomes central to the production process itself. In fact, generative tools like image generators and language models engage in their own form of audience prediction, not for the public, but for the human creator as the immediate user. In these iterative loops, the machine's imagined audience is the creator, and the goal is to approximate the creator’s expectations through each new version. This almost endless back-and-forth—try, prompt, adjust, regenerate—manifests as a form of ghostly dialogue between human and machine—where the imagined audience floats, recedes, and reappears throughout the creative process. It is simultaneously a projection, a collaborator, and a feedback mechanism, revealing the complex affective and epistemological entanglements at the heart of AI-mediated production.
Drawing on Williams’ (1974) sociotechnical framework, which positions television as a cultural form shaped by social and institutional forces, and Manovich’s (2001) new media theory, which analyzes how digital technologies alter media logics, I interrogate what qualifies as “new” in contemporary filmmaking. These frameworks help illuminate how AI and VP technologies are not merely additive tools but fundamentally restructure storytelling strategies, aesthetic values, and the hierarchy of creative roles within production ecosystems. These theoretical insights lay the groundwork for examining the implications of AI and VP not only at the level of aesthetics and narrative form but also in how labor is restructured and redistributed across global production workflows.
Building on these shifts, the labor dimension is equally crucial. Roles like script translators, localization experts, and UX designers become central intermediaries between human creators, machines, and inferred audiences. Following Soto-Sanfiel and Villegas-Simón (2024), who show how creators navigate LGBTQ+ representation under institutional pressure, this paper argues that AI-VP production also entails navigating multiple, sometimes contradictory, audience imaginaries—often with ethical consequences.
Caldwell’s (1995) concept of televisuality highlights how television evolved into a medium characterized by overt stylization, production excess, and textual density. Though developed in the context of 1980s American television, this idea helps us think about the aesthetic implications of AI-enhanced cinema. Rather than reinforcing a narrative of seamless efficiency, VP often results in an overload of information and design—aesthetic layers tuned by algorithms, real-time rendering, and predictive analytics. This paper uses Caldwell’s theory not to draw a direct parallel but to foreground how visual self-awareness and layered mediation persist, and even intensify, in AI-driven production contexts. In this way, Caldwell helps frame how imagined audiences are addressed not only through content but through the formal complexity designed to catch and retain fragmented, multitasking viewers.
In this light, the imagined audience framework makes visible the underlying power struggles—between auteurism and collective authorship, between intuition and prediction, between cultural specificity and algorithmic generalization. More importantly, the imagined audience operates as a dynamic mediator among three intersecting powers central to cultural production: the power of economic calculation, the power of aesthetic suggestion, and the power of interpretation. At the industrial level, imagined audiences are shaped by programming strategies, platform metrics, and marketing models that aim to quantify viewer behavior—reflecting the power of economic calculation. At the textual level, they are anticipated through production design, genre conventions, and narrative techniques that seek to elicit emotional resonance—demonstrating the power of aesthetic suggestion. And at the audience level, real viewers interpret content in idiosyncratic and socially contextualized ways—exercising the power of interpretation. By mediating among these forces, the imagined audience becomes a crucial site where institutional goals, creative intent, and interpretive agency converge and contend, making it a nerve center of meaning-making within AI-enhanced media production. It enables a critical understanding of how AI reconfigures authority, creativity, and meaning in contemporary media production.
In conclusion, the imagined audience offers a robust theoretical and methodological tool to analyze the transformation of creative labor and media aesthetics under AI. It highlights how AI is not just a tool but a participant in meaning-making, fundamentally entangled with how creators anticipate, negotiate, and respond to the perceived desires of others—whether human or machine. This paper contributes to rethinking media production as a space where human labor, technological systems, and aesthetic forms converge through shared, contested visions of an audience that may never be fully known but always shapes the story.
References
Andrejevic, M. (2008). Watching television without pity: The productivity of online fans. Television & New Media, 9(1), 24–46. https://doi.org/10.1177/1527476407307241
Ang, I. (1991). Desperately seeking the audience. Routledge.
Bodini, A., Manohar, A., Colecchia, F., Harrison, D., & Garaj, V. (2024). Envisioning the future of virtual production in filmmaking: A remote co-design study. Multimedia Tools and Applications, 83, 19015–19039.
Caldwell, J. T. (1995). Televisuality: Style, crisis, and authority in American television. Rutgers University Press.
Celis Bueno, C., Chow, P. S., & Popowicz, A. (2024). Not "what", but "where is creativity?": Towards a relational-materialist approach to generative AI. AI & Society. https://doi.org/10.1007/s00146-024-01921-3
Ferrari, F., van Dijck, J., & van den Bosch, A. (2023). Observe, inspect, modify: Three conditions for generative AI governance. New Media & Society. https://doi.org/10.1177/14614448231214811
Gans, H. J. (1957). The creator-audience relationship in the mass media. In B. Rosenberg & D. M. White (Eds.), Mass culture: The popular arts in America (pp. 315–321). Free Press.
Guzman, A. L., & Lewis, S. C. (2024). What generative AI means for media industries. Emerging Media. https://doi.org/10.1177/27523543241289239
Lin, B. (2024). The AI chatbot always flirts with me, should I flirt back? From the McDonaldization of friendship to the robotization of love. Social Media + Society, 10(4). https://doi.org/10.1177/20563051241296229
Litt, E. (2012). Knock, knock. Who’s there? The imagined audience. Journal of Broadcasting & Electronic Media, 56(3), 330–345. https://doi.org/10.1080/08838151.2012.705195
Livingstone, T. (2024). Game engines: Optimising VFX, reshaping visual media. NECSUS European Journal of Media Studies, Spring 2024.
Lobato, R., & Lotz, A. D. (2020). Imagining global video: Netflix. Journal of Cinema and Media Studies, 59(3), 132–136. https://doi.org/10.1353/cj.2020.0025
Manovich, L. (2001). The language of new media. MIT Press.
Soto-Sanfiel, M. T., & Villegas-Simón, I. (2024). Scriptwriters’ conceptions of audience attitudes toward LGBTQ+ characters. Mass Communication and Society, 27(5), 1252–1276. https://doi.org/10.1080/15205436.2023.2292123
Turow, J. (2011). The daily you: How the new advertising industry is defining your identity and your worth. Yale University Press.
Turow, J., & Draper, N. (2014). Industry conceptions of audience in the digital space: A research agenda. Cultural Studies, 28(4), 643–656. https://doi.org/10.1080/09502386.2014.888929
van Dijck, J. (2009). Users like you? Theorizing agency in user-generated content. Media, Culture & Society, 31(1), 41–58. https://doi.org/10.1177/0163443708098245
Williams, R. (1974). Television: Technology and cultural form. Schocken Books.
| Original language | English |
|---|---|
| Publication status | Accepted/In press - 18 Jul 2025 |
| Event | International Association for Media and Communication Research (IAMCR) 2025 Post Conference: Global Perspectives on Methods, Approaches, and Challenges of Researching AI and Labour in Creative Media Ecologies - Nanyang Technological University, Singapore, Singapore Duration: 18 Jul 2025 → … https://iamcr.org/singapore2025/methods-ai-research-conf |
Conference
| Conference | International Association for Media and Communication Research (IAMCR) 2025 Post Conference |
|---|---|
| Abbreviated title | IAMCR |
| Country/Territory | Singapore |
| City | Singapore |
| Period | 18/07/25 → … |
| Internet address |
Keywords
- Imagined Audience
- AI media production
- Virtual production