Virtual anchors and hosts on the rise
By Ye Zi
Hangzhou News Broadcast, operated by east China’s Zhejiang province’s Hangzhou Culture, Radio and Television Group, has integrated AI-synthesized virtual anchors into its programming—a pilot initiative reporting zero operational errors during broadcasts, provoking robust public debate on automation’s role in journalism.
A “virtual human” denotes a digitally engineered entity replicating anthropomorphic traits within non-corporeal environments. Accelerated by advancements in 5G networks, artificial intelligence, and immersive simulation technologies, these constructs now exhibit heightened perceptual and analytical capabilities, permeating quotidian societal functions with growing sophistication.
Hangzhou News Broadcast has deployed six such synthetic news presenters, engineered with uncanny human verisimilitude in facial micro-expressions and kinetic fluidity, according to disclosed technical specifications.
Following an intensive data acquisition and algorithmic calibration phase, these synthetic anchors now execute live news delivery with operational reliability, enabling human counterparts to temporarily step down during peak leave periods without disrupting output—a milestone in automated broadcast logistics.
Liu Yuchen, an anchor at Hangzhou News Broadcast, disclosed that her digital simulacrum “Xiaoyu” employs DeepSeek-V3 architecture, endowing it with multimodal capacities spanning natural language processing, teleprompter narration, and automated editorial functions including script vetting and source aggregation.
Media analysts note virtual presenters have transitioned from experimental applications in streaming platforms and digital-first media to institutionalized roles within terrestrial television’s daily programming cadence.
Wu Suoning, a member of the expert advisory committee of the Internet Society of China, observed that AI-driven anthropomorphic systems now constitute a transformative vector within China’s media ecosystems, operationalizing what industry protocols term a “bifurcated human-AI workflow paradigm.”
Under this framework, articulated Wu, automated precision in standardized content generation—spanning meteorological bulletins and rapid incident dispatches—allows journalistic resources to be strategically allocated to interpretive domains: investigative narrative construction and critical analysis.
The analyst projected that evolutionary leaps in large language models’ cognitive architectures, particularly in inferential reasoning and self-optimizing operational parameters, will catalyze synthetic correspondents’ capacity for adaptive interactivity and -specific customization. Such advancements, he contended, are poised to reconfigure content supply chains through hyperdynamic production protocols and precision-targeted dissemination matrices.
Beyond conventional broadcast domains, synthetic media entities are gaining traction within China’s real-time digital commerce ecosystems.
During seasonal demand peaks such as the Spring Festival, brands increasingly deploy anthropomorphic avatars — modeled after commercial ambassadors — to circumvent operational bottlenecks. Industry analyses indicate scarcity and cost inflation among skilled human anchors during high-volume periods, prompting adoption of algorithmic alternatives. These AI-driven proxies reportedly reduce expenditure while sustaining 24/7 operational cadences, mitigating fatigue-related inefficiencies inherent to biological labor.
Integrated neural language frameworks enable real-time semantic parsing, automating customer query resolution while optimizing engagement through predictive interaction protocols. Metrics from recent commercialization trials reveal conversion metrics (CVR) surging by ≥50% during AI-anchored livestreams, validating the “low-input/high-yield” operational paradigm.
Technical specifications from Baidu-affiliated digital anthropomorphism developers outline a streamlined synthesis pipeline: biometric training via 3-40 minute source footage, followed by script automation and kinematic synchronization protocols. Post-configuration, avatars achieve audiovisual congruence in lip articulation, prosody, and gesture—approximating human-presented broadcasts at scale.
Industry diagnosticians note that while synthetic presenters demonstrate superior cost efficiencies and creative scalability, they remain deficient in the nuanced adaptability of organic interlocutors—particularly in managing intricate discursive exchanges. Industry diagnosticians stress that such synthetic systems still require extensive algorithmic refinements. Concurrently, emergent debates surrounding biometric data governance, surveillance capitalism implications, and anthropomorphic ethics frameworks persist as critical impediments to sectoral maturation.
Currently, avatar personas for entertainment, real-time retail narration, and corporate service automation constitute the most monetizable implementations of anthropomorphic AI. Analytical projections from the inaugural China Digital Human Conference indicate the nation’s core virtual human sector will attain 48.06 billion yuan ($6.6 billion) by 2025, with peripheral ecosystem valuations forecast to expand to 640.27 billion yuan—underscoring accelerated enterprise adoption despite extant technical constraints.
