Daejeon, July 27–31
In March 2023, VLAST debuted PLAVE, a five-member virtual idol group that has rapidly destabilized the ontological boundaries of the K-pop industry. Powered by real-time motion capture and rendered through a game engine, PLAVE functions operationally like any other idol group—singing, dancing, engaging in livestreams—yet their bodies remain occluded by manhwa-styled avatars. To their fans (PLLIs), PLAVE represents a heightening of emotional sincerity; to broader English-language audiences, they are frequently mischaracterized as "AI idols," a label implying automation, fakeness, and the erasure of human labor.
This paper argues that PLAVE is not a rupture of K-pop's logic but a crystallization of its dependence on affective labor over bodily visibility. By triangulating multilingual topic modeling of 154,000 YouTube comments with digital ethnography, this study demonstrates how fans recalibrate authenticity through glitch circulation, defensive labor, and narrative immersion, effectively re-humanizing the avatar. Situated within the framework of techno-orientalism, the study reveals how anxieties regarding Asian technological innovation render PLAVE illegible to audiences outside Korea and Southeast Asia, positioning the group as a critical site for understanding the racialized politics of digital embodiment.
PLAVE's reception cannot be understood without addressing "techno-orientalism"—a discourse that conflates Asian identity with hyper-technological efficacy, often stripping the subject of humanity. Building on Morley and Robins (1995), who argue that Western anxieties frame Asian progress as threatening or robotic, and extending Nakamura's (2002: 6) "cybertypes," this research examines how digital avatars propagate racialized images even in virtual spaces. Miyake (2023) and Lin (2023) identify a "plastic Orientalism" or "digital-Orientalism," where Asian virtual bodies are consumed either as hyper-capitalist commodities or culturally essentialized symbols. PLAVE complicates these binaries: they neither perform the "self-orientalist" exoticism of influencers like Ling (Luo and Kim 2024) nor inhabit a dehumanized cyberpunk dystopia.
Instead, they occupy a space of meta-authenticity—a concept Kim (2025) defines as the co-construction of reality between human and technological actors, where authenticity is verified not by physical presence but by consistency of social interaction. This paper positions meta-authenticity in dialogue with idology scholarship on the "cybernetic idol" (Kovacic 2021; Nozawa 2021), arguing that PLAVE's case extends this lineage by showing how fan labor actively produces authenticity through the interpretation of technical seams. Building on Pande (2018), this reparative labor cannot be separated from race: the "re-humanization" fans perform is a response to a racialized gaze that expects Asian digital bodies to be either fully legible as Korean or threateningly artificial. PLAVE's representational opacity—Korean in voice and humor, abstract in visual form—generates both domestic affective investment and the international crisis of reception.
To empirically analyze how these theoretical tensions play out in audience reception, this project employs a computational digital humanities pipeline combined with qualitative internet ethnography.
Using Communalytic, 154,000 comments were collected from a curated corpus of 20 PLAVE-related YouTube videos (March 2023–March 2025). Videos were selected according to three criteria: (1) representation across three content categories (five Music Videos, ten Variety Shows, five Live Performances); (2) an engagement threshold of at least 70,000 views; (3) inclusion of both in-house PLAVE channel content and third-party productions, to capture cross-contextual reception. Because PLAVE has not appeared in traditional television broadcasts, YouTube constitutes the principal platform of their public exposure.
A critical methodological intervention was the refusal to treat the internet as monolingual. Using langdetect, comments were classified into Korean (74%), English (13%), Indonesian (1%), and other languages. Separate preprocessing and modeling pipelines were applied to Korean and English rather than collapsing them into a single multilingual model. This study acknowledges that "English-speaking audience" is not coterminous with "Western audience," and non-English, non-Korean corpora remain a direction for future work.
The study employs BERTopic (Grootendorst 2022), which uses sentence embeddings to cluster comments by semantic similarity rather than lexical co-occurrence. Two separate models were trained: all-MiniLM-L6-v2 for English and snunlp/KR-SBERT-V40K-klueNLI-augSTS for Korean. For each video, the top 10 non-noise topics were retained and characterized by their most distinctive keywords and representative comments, selected by proximity to the topic centroid and manually reviewed for interpretive relevance.
Computational findings were triangulated with an internet ethnography of Twitter/X and YouTube (March 2023–March 2025), following Cho's (2020) framework that treats social media comments not as static texts but as situated social interactions. Observations were structured around pre-defined analytic categories: (1) fan interactions around the "AI idol" label, (2) viral glitch moments and their framing, (3) defensive commentary and its circulation, (4) cross-linguistic interactions on shared posts.
Topic modeling reveals a stark contrast in how audiences process PLAVE's virtuality. In the English corpus of Killing Voice (PLAVE's most-viewed live performance at 4.8M views), the dominant topic centers on virtual, real, AI, motion capture, idol, with representative comments asking technical questions or correcting hostile framings ("They are PLAVE. Virtual idol. Not 'this thing'"). This indicates a visual drift where viewers struggle to anchor the avatar to a familiar performance model. Yet this skepticism is bridged by adjacent topics ( talented, deserve, voices, singing, sirens), where the sonic reality of the vocal performance overrides visual ambiguity. In the Korean corpus, discourse shifts from ontology to logistics: keywords focus on technical infrastructure (전광판 LED screen, 모션캡쳐 motion capture, 광학식 optical), suggesting a fandom that has already accepted the premise and is maintaining the technical reality of the performance. Fans prioritize emotional legibility over ontological certainty; belief is no longer anchored in sight, but in shared affective structure.
A significant finding challenges the assumption that technical failures break immersion. Ethnographic analysis of viral glitch compilations (e.g., PLAVE Glitching for 10 minutes) shows that fans actively celebrate these errors. Drawing on Park and Han (2024), this paper identifies these seams as moments of "technological vulnerability" that paradoxically heighten authenticity. Rather than exposing the idol as fake, the glitch reveals the human labor operating the real-time system. Fans have personified the technical staff—nicknamed 뚝스 Ttooks—as characters in the group's lore, integrating software developers into the parasocial relationship. PLAVE's authenticity relies on the visibility of mediation, not its concealment—a finding that complicates the "cybernetic idol" frame (Kovacic 2021) by showing how the cybernetic assemblage itself becomes an object of fan recognition.
The most quantitatively striking finding concerns the scale of fan defensive labor. In the Korean Killing Voice corpus, the single most frequent topic across all 22,000+ comments is dedicated to meta-discourse on hostile commenters (악플 akpeul), comprising 5.20% of the corpus (1,162 comments). No other cluster—including celebratory or technical topics—approaches this volume. Keywords (악플 hate comments, 굳이 going out of one's way, 답글 reply) and representative comments reveal fans systematically countering negative framings, calling out algorithmic amplification of hate, and distinguishing "real" K-pop critics from trolls. This pattern is the structural mode through which PLAVE's legitimacy is publicly maintained in Korean-language space. Paired with the Billboard Korea generative-AI controversy of April 2025—in which fans forced an official apology after identifying AI-generated imagery in a PLAVE-dedicated issue—this demonstrates the intensity with which the fandom polices the boundary between "virtual tools" (which enable human performance) and "generative AI" (which replaces it).
The study uncovers a "semiotic gap" that shapes PLAVE's reception across linguistic contexts. While VLAST CEO Lee Sung-gu has publicly articulated ambitions of Western expansion, the data shows engagement concentrated in Korea, Indonesia, and Thailand—with English-language comments representing only 13% of the corpus despite being the lingua franca of global K-pop. PLAVE defies the techno-orientalist templates through which Asian digital culture is frequently consumed: they are "too Korean" in humor and linguistic wordplay to be culturally neutral, yet "too abstract" via avatars to satisfy demands for racialized authenticity. English-language commenters frequently misidentify the group as AI—mirroring broader framings of Asian technological innovation as automated or deceptive (Nguyen and Hekman 2022). PLAVE functions as a cybertype that fails to perform the expected racial scripts, generating friction in receptions shaped by expectations of legibility.
This research contributes to the digital humanities by applying a rigorous, multilingual computational framework to the study of non-Western digital culture, challenging Eurocentric defaults in platform studies. By tracing fan practices of repairing and re-humanizing the avatar through interpretation, defensive labor, and interpretive fluency, this paper argues that PLAVE is not a novelty but a diagnostic mirror of the contemporary media condition. K-pop's authenticity has always been constructed through belief and ritual rather than physical proximity. In an era of increasing virtuality, humanity is not erased by the machine but collaboratively sought within the code.