AI 影视产业全景图 · 一分钟被观看的影像 · 2026-07 · 长内容×短剧双轨Film & TV × AI · One Watched Minute · Jul 2026 · The Long-form and Micro-drama Twin Tracks
一分钟被观看的影像
One watched minute
最小单元是一分钟被观看的影像——154 亿的院线巨制和投流三天回本的竖屏爽剧,竞争终点都是「让观众多看这一分钟」。AI 不是均匀渗透,而是沿「容错率高、非核心叙事、可标准化」的缝隙先行:后期与短剧已被吃透,宣发与素材正在被啃,立项、明星、审查是啃不动的硬骨头
The atom is one watched minute of moving image — the ¥15.4B blockbuster and the vertical drama that repays its ad spend in three days compete for the same thing: one more watched minute. AI does not soak in evenly; it advances along the seams of high fault-tolerance, non-core narrative and standardisability: post-production and micro-drama are eaten through, marketing and stock footage are being chewed, while greenlighting, stardom and censorship stay unchewable
本图相信:「吃法顺序」是理解全局的时间轴——广告营销片(2024 最先规模化)→素材库(2023 建基建)→片头空镜(个案即反弹)→正片定点特效(单镜头+高管亲自公关定调)→AI 短剧集(2026 起步)→叙事长片全流程(尚未跑通);渗透深度与「离表演和叙事的距离」严格成反比。This map believes: the «eating order» is the timeline that explains everything — ads (first at scale, 2024) → stock libraries (infrastructure, 2023) → title sequences (backlash per instance) → pinpoint VFX in features (single shots, executives personally spinning «assist not replace») → AI mini-series (from 2026) → full-pipeline narrative features (not yet achieved); depth of penetration runs strictly inverse to distance from performance and narrative.
本图不相信:「AI 将重写影视」的整句。OpenAI 的 Sora 2026-03 宣布分阶段关停——单日运营成本约 100 万美元而累计营收仅约 210 万,叠加版权深伪争议与人才经纪施压,连带迪士尼×OpenAI 逾 10 亿美元的三年合作瓦解(B)。AI 当「工具」活得好,当「主角」仍会摔跤——这与短剧侧 AI 降本增效的成功互为镜像。It does not believe the full sentence «AI will rewrite film». OpenAI announced the phased shutdown of Sora in March 2026 — roughly $1M daily running cost against ~$2.1M cumulative revenue, compounded by copyright and deepfake fights and talent-agency pressure, collapsing Disney’s $1B+ three-year deal with OpenAI (B). AI thrives as the tool and still stumbles as the star — the mirror image of micro-drama’s cost-cutting success.
主脊:一部作品的生命周期七环节(立项→融资对赌→筹备拍摄→后期→发行宣发→短剧平行轨→制度层),④⑥标红;三条结构带:吃法顺序时间轴、短剧解剖(影视业的拼多多)、制度线(劳资判例×分身租赁×审查档期);判断层:立项即命运,剧本×选角 ＞ 制作规模——AI 是效率革命,不是创作革命。姊妹:传媒→media,演出→show,IP 上游→novel,声音经济→voice,游戏→game。
The spine: one work’s lifecycle in seven links (greenlight → financing bets → prep and shoot → post → distribution and marketing → the micro-drama parallel track → the institutional layer), ④⑥ flagged; three bands: the eating-order timeline, the micro-drama dissection (film’s answer to bargain e-commerce), the institutional line (labour precedents × replica leasing × censorship and scheduling); the judgment: greenlight is destiny — script × casting ＞ production scale; AI is an efficiency revolution, not a creative one. Siblings: media → media, live shows → show, the IP headwater → novel, the voice economy → voice, games → game.
504 亿元
2024 年中国微短剧市场规模(艾媒口径)——首次超过全年电影票房;2025 年两口径打架:艾媒预测 677.9 亿 vs 广电总局口径「破千亿、较上年翻番」,相差约 350 亿必须并列(B/A)China’s 2024 micro-drama market (one research basis) — first year past the annual box office; 2025 splits between two bases: a ¥67.8B forecast vs the regulator’s «past ¥100B, doubled» — a ~¥35B gap that must be shown side by side (B/A)
148/118 天
好莱坞编剧/演员双罢工时长(2023)——全球内容业的 AI 判例:「AI 不是编剧、AI 材料不算文学材料」;数字复制品须「知情同意+公平补偿」;游戏配音罢工再打近 11 个月确立按场次付费(A)The writers’ and actors’ strikes (2023) — the content industry’s AI precedents: «AI is not a writer; AI material is not literary material»; digital replicas demand informed consent plus fair pay; the game-voice strike ran another ~11 months to set per-session fees (A)
12 小时
AI 译制一部短剧的周期(约 3000 元)vs 人工 7–15 天、1.3–7.3 万元——效率 10 倍+;出海内容中译制剧占 80–90%;AI 换脸本地化 3–10 万/部再降 20% 成本(B)——短剧是全球最激进的 AI 影像试验场AI dubbing turnaround for one drama (~¥3,000) vs 7–15 days and ¥13–73k by hand — 10×+; dubbed shows are 80–90% of exports; face-swap localisation at ¥30–100k a title cuts another 20% (B) — micro-drama is the world’s most aggressive AI-imaging test site
24.4 万元
出海短剧霸主母公司 2023 年的净利润(营收 6.86 亿、+87%,总负债约 30 亿)——买量几乎吃掉全部利润:短剧把电影的保底对赌压缩成每天用 ROI 实时清算的投流赌局(A/B)The overseas leader’s parent booked this net profit in 2023 (¥686M revenue, +87%; ~¥3B total liabilities) — ad-buying ate nearly all of it: micro-drama compresses cinema’s guarantee bets into a daily ROI-settled traffic gamble (A/B)
⚠️ 口径裁判(先读):① 短剧市场规模两口径相差约 350 亿(统计边界不同),本页并列不采信单方;② 「2026Q1 AI 微短剧占比超 95%」的报道量级远超备案口径(年约 3.3 万部),几乎肯定混入大量 UGC 生成短片,仅作趋势信号、不作产业数据;③ 厂商自述未经审计(剧本预测「86% 准确率」、各类效率提升数字);④ 一巨制的全球票房美元换算存在区间(21.2–22.6 亿美元);⑤ 2026 年事件(视频产品关停、合作瓦解、AI 短剧集)时效性强、可能反转,已保留出处;⑥ 纠偏:几家头部视频生成公司均为独立创业公司,只是与大厂有投资或合作,并非其子公司;⑦ 「春节档 200 亿」为含预售的非官方口径;⑧ 四份深度研究交叉整理(一份带源核验为主力);渗透%为编辑估值。
⚠️ Basis rulings (read first): ① the two micro-drama market bases differ by ~¥35B (different statistical boundaries) — juxtaposed, neither adopted; ② the «95%+ of Q1-2026 micro-dramas are AI» report far exceeds the registration basis (~33k titles a year) and almost certainly counts mass UGC clips — a trend signal only, not industry data; ③ vendor self-reports are unaudited (the «86% accuracy» script predictor, assorted efficiency claims); ④ one blockbuster’s global gross converts to a $2.12–2.26B range; ⑤ the 2026 events (the video product’s shutdown, the collapsed deal, the AI mini-series) are fast-moving and reversible — sources retained; ⑥ correction: the leading video-generation startups are independent companies with big-tech investors or partners, not subsidiaries; ⑦ the «¥20B Spring Festival» figure is an unofficial presales-inclusive basis; ⑧ cross-compiled from four deep-research reports (one URL-verified as backbone); penetration %s are editorial.