AI Drama \"Journey to the West: The Sequel\" Tops Ratings on Satellite TV: A Deep Dive from Technical Architecture to Industrial Revolution
I. Introduction: How a TV Series With No Actors Triggered a $1.8 Billion Market Surge?
On August 31, 2026, at 8:00 PM Beijing time, Hunan Satellite TV broadcast a “strange” television series — 30 episodes, 40 minutes each, with no human actors, no live-action filming, and every frame generated entirely by AI. This AI-produced drama, Journey to the West: The Sequel (后西游记), achieved the highest real-time ratings among all provincial satellite TV channels, with a 0.3384% rating and 1.7862% market share, generating 66 trending topics across Chinese social media.
The capital market’s reaction was even more astonishing. Mango Excellent Media hit the 20% daily trading limit for two consecutive days, with its market cap surging by over 13.2 billion RMB (~$1.8 billion). Other AI entertainment stocks followed — Huanrui Century, Huace Film & TV, and Chinese All Digital all saw significant gains.
The stock surge wasn’t driven by content quality. Rather, capital recognized a more fundamental signal — AI long-form drama industrialization had gone from zero to one. This is the world’s first AI-produced long-form drama to air on a mainstream satellite TV channel, and the first to implement the “produce-while-reviewing-while-airing” model under China’s “广电21条” policy framework.
II. Technical Analysis: A Three-Layer Industrial Production System
2.1 Overall Architecture
The technical implementation relies on a three-layer collaborative system: the model ensures consistency within a single shot; the platform ensures consistency across shots and episodes; humans ensure artistic expression quality. This three-way collaboration forms the technical foundation for 30 episodes × 40 minutes of stable output.
┌─────────────────────────────────────────────────────────────────┐
│ AI Long-Form Drama Industrial Production System │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────── Top: Human-AI Collaboration ────────────┐ │
│ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌─────────┐ │ │
│ │ │ Creative │ │ Prod. │ │ Art │ │ Visual │ │ │
│ │ │ Group │ │ Group │ │ Group │ │ Gen Grp │ │ │
│ │ │ Writers/ │ │ 3 People │ │ Scene/ │ │ ~70 Ppl │ │ │
│ │ │ Director │ │ Full Pipe│ │ Character│ │ QC/Filter│ │ │
│ │ └──────────┘ └──────────┘ └──────────┘ └─────────┘ │ │
│ └─────────────────────────┬─────────────────────────────────┘ │
│ │ │
│ ┌─────────────────────────┴─────────────────────────────────┐ │
│ │ Middle: Mango Lingchuang AIGC Platform │ │
│ │ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │ │
│ │ │Asset Lib│ │Skeleton │ │Virtual │ │Doodle │ │ │
│ │ │109 Char │ │Control │ │Camera │ │Edit │ │ │
│ │ │143 Scene│ │Anti-Clip│ │Focal/DoF│ │Local Fix│ │ │
│ │ └─────────┘ └─────────┘ └─────────┘ └─────────┘ │ │
│ │ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │ │
│ │ │Shot │ │Cost │ │Cloud │ │Model │ │ │
│ │ │Tracking │ │Tracking │ │Collab │ │Gateway │ │ │
│ │ └─────────┘ └─────────┘ └─────────┘ └─────────┘ │ │
│ └─────────────────────────┬─────────────────────────────────┘ │
│ │ │
│ ┌─────────────────────────┴─────────────────────────────────┐ │
│ │ Bottom: Seedance 2.0 / 2.5 Video Gen Models │ │
│ │ ┌─────────────────────────────────────────────────────┐ │ │
│ │ │ Dual-branch DiT: Visual Stream ◄──► Audio Stream │ │ │
│ │ │ Cross-Modal Attention Bridge in Unified Latent │ │ │
│ │ │ 30-sec gen · Native 4K · 10-bit · 50-way ref │ │ │
│ │ └─────────────────────────────────────────────────────┘ │ │
│ └───────────────────────────────────────────────────────────┘ │
│ │
│ Output: 30 episodes × 40 min = 1,200 min of AI content │
│ Assets: 109 characters + 143 scenes (reusable) │
└─────────────────────────────────────────────────────────────────┘
2.2 Foundation Model Layer: Seedance 2.0 / 2.5
The core video generation is powered by ByteDance’s Volcano Engine Seedance models, employing a Dual-branch Diffusion Transformer (Dual-branch DiT) architecture. Seedance 2.5’s key breakthrough is 30-second single-shot generation — unprecedented in the industry (Sora 2: 20s, Veo 3: 8s, Keling 3.0: 10s).
┌──────────────────────────────────────────────────────────────────┐
│ Seedance 2.0 vs 2.5 Technical Comparison │
├──────────────┬─────────────────────┬─────────────────────────────┤
│ Dimension │ Seedance 2.0 │ Seedance 2.5 │
├──────────────┼─────────────────────┼─────────────────────────────┤
│ Max Duration │ 15 seconds │ 30 seconds │
│ Max Ref Input│ 15 items │ 50 items │
│ Max Resolution│ Up to 2K │ Native 4K │
│ Color Depth │ 8-bit │ 10-bit │
│ Image Ref │ Up to 9 images │ Up to 30 images │
│ Video Ref │ Up to 3 videos │ Up to 10 videos │
│ Audio Ref │ Up to 3 audios │ Up to 10 audios │
│ Timestamp Edit│ Basic │ Integer-second precision │
│ Aspect Ratio │ 6 fixed options │ [0.4, 2.5] any ratio │
└──────────────┴─────────────────────┴─────────────────────────────┘
The 30-second threshold is transformative — it covers the “golden duration” of TikTok/TVC ads (15-30s), enabling one-shot delivery without post-production stitching.
2.3 Middle Platform Layer: Mango Lingchuang AIGC Platform
Mango TV’s self-developed cloud platform integrates 40+ image/video models and 80+ atomic capabilities. Key features include:
- Skeleton Control System: Prevents limb proportion collapse and body clipping common in AI video
- Virtual Camera System: Returns focal length, position, and depth-of-field control to creators
- Doodle-Style Local Editing: Modifies specific areas without full regeneration
- Shot Tracking: Records every generation by scene/shot, making waste rates quantifiable
2.4 Human-AI Collaboration: ~100 People
Led by Director Li Dongshen (historical documentary background), the team is organized into four groups. The key insight is that AI is positioned not as a tool but as a “creative partner” — a paradigm shift from traditional filmmaking where technology serves as a passive instrument. The Visual Generation Group, the largest team component, encompasses entirely new job categories such as AI directors, AIGC operators, and AI producers. These roles didn’t exist in the film industry just two years ago.
The production team achieved a remarkable efficiency ratio: a 100-person team accomplished what traditionally requires 2,000 people in live-action production. This 20:1 leverage ratio is the primary driver of cost reduction, though it also raises questions about the future of traditional film crews. The 3-person production group handling full pipeline scheduling is particularly noteworthy — in a traditional production, the same work would require dedicated departments for scheduling, logistics, location management, and talent coordination.
The cost structure breakdown reveals an important reality: computing costs account for only 20-25% of the total budget, while labor still accounts for 75-80%. This contradicts the popular narrative that “AI replaces all human labor.” In practice, AI long-form drama remains fundamentally a human-intensive creative process, with AI serving as an amplifier of human creativity rather than a replacement for it.
┌──────────────────────────────────────────────────────────────┐
│ Team Structure: ~100 People │
├──────────────────────────────────────────────────────────────┤
│ ┌──────────────┐ Director: Li Dongshen │
│ └──────┬───────┘ │
│ ┌──────┴──────────────────────────────────────┐ │
│ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ │
│ │ │Creative │ │Prod. │ │Art │ │ │
│ │ │Group │ │Group │ │Group │ │ │
│ │ │Writers/ │ │3 People │ │Scene/ │ │ │
│ │ │Directors │ │Full Pipe │ │Character │ │ │
│ │ └──────────┘ └──────────┘ └──────────┘ │ │
│ │ ┌────────────────────────────────────────┐ │ │
│ │ │ Visual Gen Group: 9 AI Directors, │ │ │
│ │ │ 2 AIGC Supervisors, 30 AI Producers │ │ │
│ │ └────────────────────────────────────────┘ │ │
│ └─────────────────────────────────────────────┘ │
│ Traditional: ~2,000 people → AI: ~100 people │
│ Computing costs: ~20-25% of total budget; labor: ~75-80% │
└──────────────────────────────────────────────────────────────┘
III. Code Implementation
3.1 Character Asset Consistency Management
The most critical technical challenge is maintaining character consistency across shots and episodes:
from dataclasses import dataclass
from typing import List, Dict
class ConsistencyLevel:
EXACT, FAILED = "exact", "failed"
@dataclass
class CharacterAsset:
char_id: str; name: str; model_hash: str; embedding: List[float]
class AssetConsistencyManager:
"""Ensures character visual consistency across shots/episodes"""
def __init__(self):
self.assets: Dict[str, CharacterAsset] = {}
def register(self, asset: CharacterAsset):
self.assets[asset.char_id] = asset
def verify(self, char_id: str, gen_embed: List[float]) -> str:
base = self.assets[char_id]
sim = sum(a*b for a,b in zip(base.embedding, gen_embed))
return ConsistencyLevel.EXACT if sim >= 0.95 else ConsistencyLevel.FAILED
# Register Sun Xiaosheng, verify cross-shot consistency
mgr = AssetConsistencyManager()
mgr.register(CharacterAsset("SXS_001","Sun Xiaosheng","sd2.5_v3",[0.92,0.87,0.45]))
print(mgr.verify("SXS_001", [0.91,0.86,0.44])) # exact
3.2 Generation Cost Tracking
AI long-form drama has “high marginal cost and high rework cost”:
package main
import "fmt"
type GenTask struct{ ID, Model string; Cost float64 }
type CostTracker struct {
cmap map[string]float64
used float64
}
func (c *CostTracker) Add(t GenTask) {
c.used += t.Cost
cat := "compute"
if t.Model != "seedance2.5" && t.Model != "seedance2.0" { cat = "labor" }
c.cmap[cat] += t.Cost
}
func main() {
ct := &CostTracker{cmap: make(map[string]float64)}
for ep := 1; ep <= 5; ep++ {
for sh := 1; sh <= 20; sh++ {
ct.Add(GenTask{fmt.Sprintf("EP%02d_SH%02d",ep,sh),"seedance2.5",250+float64(sh)*5})
}
}
fmt.Printf("Total: %.2f | Compute ratio: %.1f%%\n", ct.used, ct.cmap["compute"]/ct.used*100)
}
3.3 Version Control for “Produce-While-Review-While-Air”
from typing import Dict
class EpisodeVersion:
def __init__(self, ep: int, ver: int, content: Dict, parent: int = None):
self.ep = ep; self.ver = ver; self.content = content; self.parent = parent
class ConcurrentManager:
"""First 5 episodes air → collect feedback → dynamically adjust subsequent plots"""
def __init__(self, batch: int = 5):
self.versions: Dict[str, EpisodeVersion] = {}
for ep in range(1, batch + 1):
self.versions[f"EP{ep:02d}_V01"] = EpisodeVersion(ep, 1, {})
def adapt(self, ep: int, fb: Dict) -> EpisodeVersion:
last = max((v for k,v in self.versions.items() if k.startswith(f"EP{ep:02d}")),
key=lambda x: x.ver)
nv = EpisodeVersion(ep, last.ver+1, last.content.copy(), last.ver)
nv.content["adaptations"] = fb.get("adaptations", [])
self.versions[f"EP{ep:02d}_V{nv.ver:02d}"] = nv
return nv
pm = ConcurrentManager()
a = pm.adapt(6, {"adaptations": [{"type":"pacing","adj":"speed up celebration"}]})
print(f"Version: EP{a.ep:02d}_V{a.ver:02d}")
IV. Cost and Efficiency: ~$125K Per Episode
4.1 Cost Comparison
┌─────────────────────────────────────────────────────────────────┐
│ AI Drama vs Traditional Drama Cost Comparison │
├──────────────┬─────────────────────┬────────────────────────────┤
│ Dimension │ AI Drama │ Traditional S-Grade │
├──────────────┼─────────────────────┼────────────────────────────┤
│ Per Episode │ ~900K RMB │ Tens of millions RMB │
│ Per Minute │ 20-30K RMB │ Hundreds of thousands │
│ Prod. Cycle │ 3-4 months │ 1-2 years │
│ Team Size │ ~100 people │ ~2,000 people │
│ 3-min Fight │ 20-30K RMB │ 4-5 million RMB │
│ Char Assets │ 109 (reusable) │ Built from scratch │
│ Scene Assets │ 143 (reusable) │ Built from scratch │
└──────────────┴─────────────────────┴────────────────────────────┘
4.2 Cost-Reduction Mechanisms
Two key restructuring points drive cost reduction:
1. Skipping 3D modeling: Traditional animation requires a complete pipeline: concept art → 3D modeling → texturing → rigging → animation → rendering. Each step is labor-intensive and requires specialized expertise. AI drama bypasses the most expensive intermediate steps, generating final-quality images directly from concept art and text prompts. A single 3-minute fight sequence that would cost 4-5 million RMB using traditional VFX methods was generated for just 20-30K RMB using Seedance 2.5 — a cost reduction of over 99%.
2. Rendering paradigm shift: Traditional VFX renders frame-by-frame, with complex scenes potentially taking hours per frame at costs of thousands of RMB each. AI generates entire sequences in one pass, eliminating the frame-by-frame cost structure entirely. The show’s most visually complex scenes — the Flower-Fruit Mountain celebration with hundreds of animated characters, the underwater journey to the Dragon Palace, and the celestial battle sequences — would have been prohibitively expensive using traditional techniques. With AI, they became economically viable.
┌─────────────────────────────────────────────────────────────────┐
│ Traditional 3D: Concept→Model→Texture→Rig→Animate→Render │
│ Timeline: 1-2 years, Cost: 10x │
│ │
│ AI Drama: Concept→AI Generate→Filter→Local Repair │
│ Timeline: 3-4 months, Cost: 1/10 │
└─────────────────────────────────────────────────────────────────┘
4.3 Cost Structure
Computing costs account for 20-25% of total budget; labor accounts for 75-80%. A 100-person team accomplishes what traditionally requires 2,000 people.
┌─────────────────────────────────────────────────────────────────┐
│ Cost Structure Breakdown │
├─────────────────────────────────────────────────────────────────┤
│ ┌──────────────┐ │
│ │ Computing │ 20-25% │
│ │ (Seedance) │ │
│ └──────┬───────┘ │
│ ┌──────┴────────────────────────────────────┐ │
│ │ Labor (75-80%) │ │
│ │ ┌──────────┐ ┌──────────┐ ┌────────┐ │ │
│ │ │Creative │ │AI Prod/ │ │Prod/ │ │ │
│ │ │/Writing │ │Ops 40% │ │Mgmt 25%│ │ │
│ │ │ 15% │ │ │ │ │ │ │
│ │ └──────────┘ └──────────┘ └────────┘ │ │
│ └───────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
V. The “Produce-While-Review” Model
5.1 Policy Background
Under the “广电21条” policy, Journey to the West: The Sequel became the first drama to implement the “produce-while-reviewing-while-airing” model. The review cycle is compressed from months to weeks or even days.
┌─────────────────────────────────────────────────────────────────┐
│ "Produce-While-Review-While-Air" Workflow │
├─────────────────────────────────────────────────────────────────┤
│ Timeline: │
│ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ │
│ │Ep1-5 │→│ Air │→│Collect│→│Adjust │→│Ep6-10│ │
│ │Produce│ │Ep1-5 │ │Feedback│ │Ep6-10│ │Produce│ │
│ │Review │ │ │ │Barrage │ │ │ │ ... │ │
│ └──────┘ └──────┘ └──────┘ └──────┘ └──────┘ │
│ │
│ Traditional: Full Production (1-2yr) → Full Review → Air │
│ Concurrent: Batch Produce → Review (Days) → Air → Feedback │
└─────────────────────────────────────────────────────────────────┘
5.2 Natural Fit for AI Drama
AI content has high marginal costs — more detail means higher modification costs. The batch-review model allows rapid airing and dynamic adjustment based on audience feedback, leveraging AI’s “fast production, fast feedback, fast iteration” advantages. The production team has stated that subsequent episodes will be adjusted based on barrage comments.
VI. Market Impact: Billions in Market Cap Surge
6.1 Capital Market Reaction
┌─────────────────────────────────────────────────────────────────┐
│ Capital Market Impact Overview │
├─────────────────────────────────────────────────────────────────┤
│ Mango Excellent Media: │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ Aug 31: 20% daily limit hit (within 16 min of opening) │ │
│ │ Sep 1: 20% daily limit (second consecutive day) │ │
│ │ As of Sep 7: Cumulative gain > 50% │ │
│ │ Market cap surge: > 13.2 billion RMB (~$1.8B) │ │
│ └──────────────────────────────────────────────────────────┘ │
│ │
│ Sector-wide rally: Chinese All Digital (+20%), Rongxin (+20%), │
│ Huanrui Century (+10%), Bona Film (+10%), Huace (+15%) │
│ │
│ Key Insight: Capital is betting on AI industrialization │
│ feasibility, not the show's content quality. │
└─────────────────────────────────────────────────────────────────┘
6.2 Capital Logic
The surge was driven by concentrated validation of multiple breakthroughs:
- Industrialization: Complete end-to-end execution from concept to broadcast
- Policy window: First “produce-while-review” drama, new regulatory paradigm
- Cost benchmark: ~900K RMB/episode shows disruptive ROI potential
- User growth: Male user acquisition exceeded expectations for Mango TV
VII. Polarized Reception
7.1 Positive Reviews
- “Visual quality exceeded expectations — couldn’t tell it was AI-generated”
- “Unique Chinese mythological aesthetic with historical provenance”
- Character merchandise (plush toys) launched simultaneously on e-commerce
7.2 Critical Reviews
- “Too much ‘AI feel’ — lacks fluidity and liveliness”
- “Pacing is sluggish, plot is flat”
- “Facial expressions are stiff, lacking subtle emotional expression”
Professor Xu Miaomiao noted that AI-generated characters “look like the character but don’t feel like the character.”
7.3 Hu Xijin’s Opposition
Prominent commentator Hu Xijin published an article opposing “AI technology replacing human performance in large-scale realist dramas.” His stance sparked widespread debate — supporters see him as “speaking for ordinary people,” critics view him as “a rickshaw driver blocking automobiles 100 years ago.”
VIII. Industry Significance: Three Milestones
8.1 Technology Milestone
The most significant technical achievement of Journey to the West: The Sequel is proving that AI video generation can scale from 15-second clips to 40-minute episodes while maintaining visual consistency. This leap required solving three fundamental challenges: character identity preservation across hundreds of shots, coherent scene transitions without jarring visual shifts, and sustained narrative pacing that doesn’t rely on the “rapid-fire editing” crutch common in short-form content. The Seedance 2.5 model’s 30-second single-shot capability was critical — it enabled the creative team to generate complete scenes as unified narrative units rather than stitching together discrete clips.
8.2 Regulatory Milestone
As the first drama to implement the “produce-while-review-while-air” model under the 广电21条 framework, Journey to the West: The Sequel created a new regulatory paradigm. The traditional model required all 30 episodes to be completed and reviewed before any could air — a process taking months. The new model allows batches of episodes to be reviewed in days, with the first 5 episodes airing while episodes 6-10 are still in production. This not only accelerates time-to-market but also enables a novel feedback loop where audience reactions can shape subsequent episodes in near real-time. The policy implications extend beyond AI content — this model could eventually transform how all television content is produced and distributed.
8.3 Industry Milestone
For the first time, an AI production pipeline has been established that can be replicated and scaled. The 109 character assets and 143 scene assets are not just byproducts of a single production — they constitute a reusable digital library that can be deployed across future projects. This asset-reuse model fundamentally changes the economics of content production: the first season is expensive, but subsequent seasons benefit from compounding asset value. Mango TV has already announced plans to expand the IP into a multi-season franchise, with the “Winter Arc” currently in pre-production. The “Flower-Fruit Mountain” arc (first 5 episodes) generated over 2.1 billion views on Mango TV alone, reaching 700,000+ households through the “剧好看” big-screen platform that spans 31 provinces across 5.51 billion TV terminals.
8.4 Global Reach
Over 750 international media outlets reported (Yahoo, AP, Business Insider), reaching 220M+ overseas viewers. Korean media called it a “milestone for AI-produced television.”
IX. Challenges and Future Outlook
9.1 Current Challenges
Despite its milestone success, the challenges facing Journey to the West: The Sequel are significant. The most pressing technical limitation is emotional continuity — AI-generated characters still struggle with nuanced facial expressions and subtle emotional cues. In the show’s more dramatic scenes, characters’ faces remain relatively static, undermining the emotional impact of key narrative moments. This is particularly problematic for long-form content where audiences develop attachment to characters over 40-minute episodes.
Narrative depth presents another hurdle. While AI excels at generating visually spectacular scenes — the “Flower-Fruit Mountain” celebration sequence is genuinely stunning — it struggles with sustained narrative tension and character development. Critics noted that the plot relies too heavily on visual spectacle at the expense of story coherence, echoing the “padding” problem that plagues traditional long-form television.
Monetization remains the most existential challenge. Traditional TV economics rely heavily on star power for advertising and subscription revenue. AI-generated content has no celebrity endorsements, no fan followings, and no established brand value. The show’s initial付费 (paywall) data has been encouraging but not definitive. Mango TV’s SVIP model (early access to 3 episodes) and VIP model (early access to 2 episodes) provide a starting point, but sustainable monetization strategies remain unproven.
Legal and regulatory risks also loom large. Questions around AI training data copyright, the boundary between generated character images and real person肖像权 (portrait rights), and the ethical implications of AI replacing human performers all remain unresolved. Industry associations have called for accelerated development of legal frameworks to keep pace with technological change.
Finally, the industry ecosystem remains in its infancy. While the show’s production team successfully navigated the complexities of AI production, there are no standardized training programs, established career paths, or proven best practices for AI filmmaking. Each new project essentially rebuilds the workflow from scratch.
9.2 Future Outlook
- Technology: ByteDance is developing real-time spatial video generation (overseen by founder Zhang Yiming)
- Business models: Virtual idol operations, IP derivatives, virtual concerts
- Content quality: Moving from “technology validation” to “content excellence”
- Ecosystem: ByteDance’s Jimeng AI launched “Jimeng Studio” for film/TV projects
As Director Li Dongshen observed: “If the industry only chases short-term gains of AI short-form content, using AI merely as a cost-cutting tool, once the technology dividend window closes, only short-lived content may remain.”
X. Conclusion and References
Conclusion
The satellite TV broadcast of Journey to the West: The Sequel marks the transition of AI film/TV from “technology experiment” to “formal content production.” It achieved zero-to-one breakthroughs across three dimensions — technology, regulation, and industry — validating the feasibility of AI long-form drama industrialization. The show demonstrated that AI can generate 40-minute episodes with consistent visual quality, that regulators can adapt review processes to accommodate AI production workflows, and that a replicable industrial pipeline can be established for future projects.
However, the road ahead is far from smooth. The show’s polarized reception highlights the gap between technological capability and artistic maturity. AI can generate stunning visuals, but it has not yet mastered the subtle art of storytelling — the emotional beats, the character arcs, the narrative pacing that separates good television from great television. The challenges of emotional continuity, narrative depth, and sustainable monetization will require years of iterative improvement to resolve.
What makes the Journey to the West: The Sequel story remarkable is not that it produced a perfect show — it didn’t. What makes it remarkable is that it produced a show at all, proving that the technical and economic barriers to AI long-form drama are no longer insurmountable. The capital market’s response — a $1.8 billion surge in market value — reflects a bet on the trajectory, not the current state. As Director Li Dongshen noted, “If the industry only chases short-term gains, using AI merely as a cost-cutting tool, once the technology dividend window closes, only short-lived content may remain.”
The era of AI long-form drama has arrived. The question is no longer whether AI can produce television, but how quickly the industry can evolve to produce television that is not just technically competent, but artistically compelling.
References
- Titan Media: “How Hou Xiyou Ji Leveraged $1.8B Market Value” (2026.09.08)
- NetEase: “First AI Long-Form Drama Tops Ratings” (2026.09.01)
- Jintai News: “AI Long-Form Drama Premieres, Industry Shift” (2026.09.08)
- Red Star News: “First Satellite-Aired AI Long-Form Drama” (2026.08.31)
- Qianjiang Evening News: “First AI Drama on Hunan Satellite TV” (2026.09.01)
- Beijing Youth Daily: “AIGC Long-Form Drama Tops Ratings” (2026.09.01)
- Volcano Engine Docs: Seedance 2.0/2.5 Specs (2026.08-09)
- Sina Tech: “Hu Xijin Comments on AI Drama” (2026.09.01)
- Hunan Satellite TV: “Hou Xiyou Ji Sets New Possibilities” (2026.09.07)
- Liuhe Business Narrative: “AI Long-Form Drama: New Paradigm” (2026.09.05)