Netflix 刚刚披露，其斥资 5.87 亿美元收购了本·阿弗莱克（Ben Affleck）的 AI 创业公司。

Netflix just disclosed it paid $587 million for Ben Affleck's AI startup. Sixteen employees. Three years in stealth. That works out to $36.7 million per person, before you even try to separate the value of the technology from the value of the team. Here is what actually makes this interesting. InterPositive is not a generate-a-scene-from-a-prompt company. It fixes footage you already shot. Missing shots, inconsistent lighting, continuity errors, background replacements. Netflix co-CEO Ted Sarandos put it plainly on the earnings call: productions that could never have afforded a complicated crowd scene or a historic battle sequence can now include them, because the tools make those sequences achievable within the budget the production already has. The real money in AI filmmaking right now is not going toward tools that replace a shoot. It is going toward tools that make a shoot that already happened more ambitious than its budget would allow. The same week, a different story made the same point from the opposite direction. Christopher Nolan's Odyssey opened in IMAX with a $250 million budget. AI studio Fountain 0 released its own adaptation of the same material, 135 minutes, built by director Ash Koosha on a budget in the mid-five figures, roughly 5,000 times cheaper. Kling for video generation, Claude for language editing, twelve licensed human likenesses including Koosha's own as Odysseus. Koosha was honest about the limits: "I would never be able to get a budget to display a monster in the sea and a cyclops in a cave" at Nolan's scale. He is not claiming parity. He is claiming access to a story he could never have gotten a traditional budget to greenlight. Both stories land on the same truth. The value was never in the generation itself. It is in the plan behind it, and the specific problem someone was actually trying to solve. Also in this edition: the real workflow behind character consistency (it stopped being a prompting problem a while ago), and a quick correction on something we got wrong in the last edition. Full breakdown in Edition 12 of AI Film Radar: https://t.co/vtC67KGL3w #AIFilm #Filmmaking #AICinema #Netflix

Netflix 刚刚披露，其斥资 5.87 亿美元收购了本·阿弗莱克（Ben Affleck）的 AI 创业公司。 16 名员工。三年的隐身研发期。这样算下来人均 3670 万美元，这还没算要把技术的价值与团队的价值区分开来。 真正有意思的地方在于，InterPositive 并不是一家"输入提示词生成场景"的公司。它修复你已经拍好的素材。缺失的镜头、不一致的灯光、连贯性错误、背景替换。Netflix 联合 CEO 特德·萨兰多斯（Ted Sarandos）在财报电话会议上一针见血地指出：以前根本负担不起复杂群戏或历史战争场面的制作，如今也能加入这些场面了，因为这些工具让它们能够在制作已有预算范围内实现。 当下 AI 电影制作领域真正的钱，并没有流向那些"取代拍摄"的工具，而是流向了那些让已经完成的拍摄比其本应支撑的预算更宏大、更具野心的工具。 同一周，另一个故事从相反的方向印证了同一个观点。克里斯托弗·诺兰（Christopher Nolan）的《奥德赛》以 2.5 亿美元预算登陆 IMAX 银幕。AI 工作室 Fountain 0 则推出了对同一题材的改编版本，时长 135 分钟，由导演 Ash Koosha 制作，预算仅为五位数中段，比前者便宜约 5000 倍。视频生成用 Kling，语言编辑用 Claude，使用了十二个获得授权的人脸形象，其中包括 Koosha 本人的奥德修斯形象。 Koosha 对局限性坦率承认："以诺兰那种规模，我永远拿不到预算来在海上呈现怪物、在洞穴里呈现独眼巨人。"他并未声称能比肩诺兰，他主张的是获得一种他用传统预算永远得不到立项的故事讲述机会。 两个故事都指向同一个真相。价值从来不在生成本身，而在于其背后的策划，以及有人真正试图解决的那个具体问题。 本期内容还包括：角色一致性背后的真实工作流（这个问题早在不久前就不再是一个提示词的问题了），以及对上一期一处错误的快速更正。 详细内容请见《AI Film Radar》第 12 期：https://t.co/vtC67KGL3w #AIFilm #Filmmaking #AICinema #Netflix