电影史上，新工艺登台时从来算不上斯文。
1895 年，《火车进站》让观众惊慌躲闪。1927 年，《爵士歌手》一声开口，默片时代逐步谢幕。1995 年，《玩具总动员》甚至让观众忘了它是完全由电脑生成。
有意思的是，每一次「离经叛道」，最后都加速了行业发展。有声电影、彩色胶片、数字摄影机、CG 特效，无一不是在怀疑与嘲笑中入场，最终被写进电影语言的词典。
如今，生成式 AI 也开始参与影视创作，一家以故事创作为核心的 AI 原生影视公司 Utopai Studios，正凭借旗下的 Utopai X，搅动视频生成模型的格局。
截至 9 月 30 日，其自研模型 Utopai X 在独立实测机构 Artificial Analysis 的带音频文生视频排行榜中位列全球第二，仅次于 Wan 3.0，压过 Dreamina Seedance 2.5，同时是该榜排名最高的美国公司模型。
热知识，该榜使用 Video Arena 盲评机制：观众不知道模型品牌，仅凭观看感受为同一提示词生成的视频投票，排名由投票形成的 Elo 分数决定。这也是榜单引入盲评以来，首次有影视公司的自研模型上榜，堪称出道即巅峰。
成绩之外，更值得关注的是这家公司的身份。
业务覆盖影视项目开发、融资、制作与全球发行，模型与平台都服务于实际的电影和剧集生产。对影视从业者而言，这是一个难得的样本：生成式 AI 如何从单个镜头的创作，进入一部完整作品的制作流程。
一家影视公司，杀进了视频模型的决赛圈
Utopai 总部位于美国加利福尼亚州山景城，联合创始人 Cecilia Shen 与 Jie Yang 分别担任首席执行官和首席工艺官。Cecilia Shen 此前曾在 Google X 从事数据科学与 AI 应对思路相关工作。
Utopai 的发展始于 2022 年成立的 Cybever。
早期研发围绕程序化内容生成与高保真三维环境展开，为游戏和影视预演打地基，随后摸着石头过河，逐步延伸到空间推理、镜头调度，以及从三维预演生成电影级视频。
2025 年，Cybever 更名为 Utopai Studios，开始围绕原创作品和联合制作，建立影视开发、制作与发行体系。自研工艺也随之被放进真实制作中，用于兼容角色、情形和镜头的开发。
据《福布斯》估算，Utopai 的估值约为 10 亿美元。《综艺》在刊发的合作专题中，将其描述为「全球最大独立 AI 原生影视公司」，并介绍了计划于 2027 年推出的三部电影与两部剧集。
影视制作的实际需要，给 Utopai 的模型研发当了出题人。
要知道，用视频生成工具，使用者可以反复「抽卡」，长片制作却是另一套打法：数百甚至数千个镜头要连成一条完整的叙事线，角色、环境、道具的也得经得起摄影机来回折腾。
制作过程还伴随着大量修改。
导演可能认可了角色外观，却希望调整光线；可能保留了情形布局，却需要改变人物动作。每一次修改，都要尽量延续此前已经确定的创作挑选，让新素材可以与其他镜头一起使用。
生成质量、跨镜头一致性与可修改性，由此成了长片制作必须同时翻越的三座大山。
Utopai X 的研发，也围绕这些制作要求展开。Artificial Analysis 披露，该模型基于 MiniMax H3 开展后训练，此次排名超过了 MiniMax H3 原模型，以及同样基于它开发的 MiniMax H3 Max。
依据最新资料信息，Utopai X 在音频同步与物理表现两个能力类别中排名第一，在光照与材质、摄影机控制两个类别中排名第二。
这些能力条条对应片场的硬要求：声音要和画面动作合拍；光线穿过玻璃、水体或金属时，变化得经得起推敲；摄影机一动，角色与环境之间的空间关系不能断线。
针对反射、焦散、空间理解与高级一致性等难点，Utopai X 也一一开展优化，目的在于提高素材的可用性，让电影人可以围绕创意调整镜头，减少反复生成后重新筛选的工作。
从 Utopai X 到 PAI，让生成能力进入制作流程
模型可以生成怎样的画面，还需要借助制作体系转化为电影人可以使用的能力。
Utopai 为此开发了 PAI 制片智能平台，并将 Utopai X 深度集成其中。两者分别承担不同的工作，Utopai X 给出视频生成能力，PAI 则组织剧本、角色、情形、镜头及创作反馈，让生成过程可以沿用项目已有的信息。
在 PAI 中，剧本可以被分解为情形和镜头，角色、地点、道具与视觉参考全程关联于项目。创作者推进新镜头时，可以直接调用已确认的设定，不必每生成一次就把整个故事世界从头讲一遍。
平台内的 AI agents 以「制片助理」的身份进场，是名副实际上的神助攻：基于制作上下文帮电影人规划镜头、试创意、出多个版本。创作者则可以货比三家、回溯旧版，再拍板做什么、留什么、改什么。
选定的素材随后进入剪辑时间线，并与 Premiere、DaVinci Resolve 等专业后期工具衔接，形成从创意开发、素材生成到后期交付的连贯工作流。
整个过程中，故事、视觉风格、表演与最终版本的生杀大权，始终握在电影人手里。首席科学官 Zijian He 所强调的「工艺应该扩展电影人的画布」，也体现在这样的分工里，让创作者具备更多尝试空间，并对结果保持掌控。
研发与制作在同一屋檐下并行，近水楼台，自然转出了一条持续迭代的飞轮。
艺术家在片场踩到的坑，会直接成为研发需要处理的难题，例如，同一个角色如何跨情形保持一致，已经借助审核的构图如何在修改动作后继续保留，多个版本怎样进入剪辑并供主创比较。
实际制作攒下的数据与专家反馈，被用来打磨模型、智能体和工作流程，磨出来的新本事再回到当前项目，并像滚雪球一样带进后续作品。
相比只围绕独立片段开展测试，专业制作可以给出更完整的反馈。一个镜头是否漂亮之外，电影人还会判断它能否与上下镜头衔接、能否继续修改，以及能否满足最终交付要求。
中国 IP 与全球制作，有了新的合作方式
榜单之外，Utopai 的制作体系正在进入不同类型与文化背景的故事，接受比单个视频片段更复杂的检验。
计划于 2027 年推出的动画长片《The Most Serious Fart》，改编自 Mike Bender 于 2023 年出版的同名儿童绘本。故事围绕一个格外严肃的屁 Siegfried 展开，他希望改变外界对自己的看法，并在旅途中学习接纳自己、找到归属感。
看似荒诞的设定，承载着身份认同、自我接纳与打破偏见的主题。Utopai 希望从具有辨识度的角色和故事世界出发，开发面向全球观众、可以长期延展的娱乐 IP，让一部电影成为持续创作的起点。
主创阵容也体现了这种开发思路。
原著作者 Mike Bender 将担任编剧和导演，他此前参与过喜剧电影《Not Another Teen Movie》，也是《Awkward Family Photos》的共同作者。Russell Hollander 担任制片人，原著插画师 Chuck Dillon 则出任创意艺术总监，让绘本原有的角色气质与视觉特征延续到电影中。
参与《怪物史莱克》编剧、并凭该片取得奥斯卡最佳改编剧本提名的 Joe Stillman，担任联合执行制片人；参与《驯龙高手》等作品的 Andy Brooks 加入故事与分镜创作；曾在皮克斯担任艺术指导的 John Lee 负责制作外观。《辛普森一家》《飞出个将来》的选角导演 Scott Muller 则负责配音选角。故事、视觉与声音的关键环节，都由具有专业制作经验的创作者参与。
据 Utopai 披露，人类动画师与插画师主导创作，AI 辅助制作迭代，角色的情感与故事的节奏仍由主创把握。
同一片单中的家庭电影《Half Moon》，由德国 In Good Company、韩国 Paper Barn Studios 等伙伴共同参与，PAI 兼容部分视觉元素。历史史诗《Cortés》则面对另一种制作难度，需要让规模庞大的历史情形进入一部可以完成、可以发行的电影。
《Cortés》的故事发生于 1519 年，围绕埃尔南·科尔特斯远征阿兹特克帝国展开，涉及西班牙征服者与美洲原住民之间的冲突。奥斯卡提名编剧 Nicholas Kazan 曾表示，他为讲述这个故事努力了三十多年，宏大的规模与高昂的成本始终是项目推进的障碍。
题材的难度还来自历史本身。
西班牙人的到来、战争与特诺奇蒂特兰的陷落，涉及复杂且存在争议的历史叙述。Kazan 强调，影片希望承认不同立场与视角的存在，借助一个具体故事展现两种文化相遇与冲突的过程，避免将某一方的叙述当成全部历史。这种尽量兼顾多方视角的创作立场，也意味着主创需要谨慎处理历史人物与冲突，如何理解不同群体的处境，与如何重建古城和战争场面同样关键。
把特诺奇蒂特兰古城、战争与帝国变迁搬上银幕，意味着复杂的情形构建、群像调度和视觉制作。Utopai 计划借助 AI 制作工具缓解这些规模与成本限制，为这个长期未能进入银幕的故事给出新的制作条件。
影片由 Paul Mignot 执导、Marco Weber 制片，以巴黎为制作基地，在西班牙拍摄，目标于 2027 年上映。编剧持续三十多年的尝试，与 Utopai 在制作工艺上的投入，开始指向同一个结果，让故事成为观众可以观看的电影。
Utopai 在亚洲的合作同样也是量体裁衣，按作品的叙事需要来挑选合适的工艺。
Utopai 与华策影视共同开发并发行动画剧集《西游记：消失的 500 年》，将 PAI 用于开发与制作；与日本 DeNA 的合作覆盖视觉概念、角色、环境及情形迭代，创意审核权仍由 DeNA 艺术家掌握。
韩国业务则以 Alquimista Media 的创意团队为基础，亚太负责人 Hyun Park 更是具备 Studio Dragon、华纳兄弟韩国及 DramaFever 的从业经历。
AI 可以帮创作者更快试画面，人物、文化语境和叙事取舍，仍需要熟悉作品的人来把关。
从好莱坞动画到中国经典 IP 这些合作，笔者看到的并不是一个 AI 如何颠覆影视工业的故事，这些尝试让制作工艺适应各地的审美与叙事，把本地创意和跨市面的制作、发行经验捏合到一起，借船出海。
这或许是 AI 和影视两个行业更健康的合作模式，这类作品的发行方式也在改变。
与 Ex Machina Studios 等伙伴合作的《Space Nation》，是一部计划推出八集的长篇科幻剧集，围绕人类面对外星威胁展开。对 PAI 而言，连续剧提出的要求远远超过生成几段太空画面，角色、环境、视觉风格与叙事衔接，需要在整季作品中持续成立。
从世界观与角色开发，到不同环境的视觉构建，再到多集内容之间的连贯性维护，长篇科幻作品会反复检验制作工具能否在复杂项目中持续使用。一个可以长期发展的系列 IP，还需要让后续故事沿用已有设定，避免每次制作都重新建立整个世界。
《Space Nation》也成为首部与传统版权买家达成合作的 AI 剧集，巴西 Globoplay、德国电视台等买家按非 AI 内容的市面售价采购了相关发行权。
据《福布斯》报道，在最终作品按约交付的前提下，《Space Nation》与《Cortés》的潜在合计预售收入最高可达约 1.1 亿美元。
版权交易背后，两个项目的推进远没有想象中那么简单。
《Cortés》与《Space Nation》早在 2025 年就已传出消息，如今才朝着计划于 2027 年推出的目标继续前行。究其原因，漫长的推进过程，正是发生在好莱坞对 AI 使用边界格外敏感的时期。
据《福布斯》报道，《Cortés》计划使用真人表演与数字环境结合的方式，拍摄演员的表演，再将其放入数字情形。而这类项目所需人力要远少于同等规模的传统制作所需。对于制片方，这是让昂贵故事有机会启动的条件；对于从业者，同一组数字也意味着工作机会可能发生变化。
争议因而有了具体对象。真人拍摄结束后，演员的形象可以被修改到什么程度，能否用于其他镜头甚至后续作品，授权与报酬怎样计算，都直接关系到这种制作方式能否被接受。
2023 年罢工后，SAG-AFTRA 达成的协议已经为数字替身设置了知情同意与报酬要求，保护范围同时覆盖主演和群演，制作方不能把一次拍摄理解为无限使用演员形象的许可。
编剧同样争取到了明确边界。美国编剧工会（WGA）的 2023 年协议规定，制作方不能强迫编剧使用 AI，也不能利用 AI 生成材料削弱编剧的署名与相关权益。对于把 AI 深度引入制作的工作室，保留人类创作者的决定权，需要落实到剧本开发、素材使用与合同安排之中。
面对行业疑虑，Cecilia Shen 在《福布斯》采访中强调，Utopai 使用不侵犯版权的训练数据，愿意与行业工会合作，并将创意决策保留在人类手中。这些是 Utopai 对外作出的回应，尚不能据此认定双方已经应对所有分歧。
到了 2026 年，《综艺》合作专题披露，《Space Nation》已进入制作，并计划按照好莱坞工会框架推进。从公开项目到计划发行，Utopai 的坚持，也体现在愿意为工艺之外的协商与制作投入时间，每一次向前推进，都意味着同时处理创作、预算、制作与行业规则的新的突破。
而且 Utopai 还是首家加入好莱坞编剧协会与演员工会的 AI 影视公司，算是拿到了传统影视圈的正式「户口」。电影节、专业主创、行业规则与版权交易，一步步把它迎进了产业的正门，「AI 原生影视公司」这块招牌，也开始对应实打实的制作与发行。
新工具很少杀死旧艺术，但它会重新划定讲故事的门槛。就像摄影浮现后，绘画没有退场，反而让画家则有了更多理由探索镜头拍不到的东西，摄影也逐步发展成新的艺术语言。
AI 视频今天正走在这条老路上。Utopai 也已经拿到了入场券，它把模型研发放进了影视生产需要承担的责任之中。接下来，这些作品能否让观众追下去、更多曾被预算挡在门外的故事能否落地，才是这套模型与平台的真正考验。
大银幕的感受不能完全用 Elo 分数来衡量，能让更多过去因制作条件被束之高阁的故事破土而出、登上银幕，并让观众愿意坐到字幕升起，AI 才真正从演示视频走进了电影。
本文来源：https://www.ifanr.com/1682888
若有版权问题，请来信告知：13951431913#139.com 或在右方提交评论。
In film history, new technology has never exactly arrived politely.
In 1895, L'Arrivee d'un Train en Gare sent audiences fleeing in panic. In 1927, one spoken line from The Jazz Singer gradually ended the silent era. In 1995, Toy Story even left audiences forgetting it was entirely computer-generated.
What is interesting is that every such act of 'defying convention' ultimately accelerated the industry's development. Sound films, colour film, digital cameras, CG effects — none of them entered without doubt and mockery, yet all were ultimately written into the dictionary of cinematic language.
Now generative AI is also beginning to participate in film and television creation, and an AI-native studio centred on storytelling, Utopai Studios, is shaking up the landscape of video generation models with its Utopai X.
As of September 30, its in-house model Utopai X ranks second globally on Artificial Analysis's text-to-video-with-audio leaderboard, behind only Wan 3.0 and ahead of Dreamina Seedance 2.5. It is also the highest-ranked US company model on that leaderboard.
Worth knowing: that leaderboard uses the Video Arena blind evaluation mechanism — viewers do not know the model brand and vote purely on viewing experience for videos generated from the same prompt, with rankings determined by the Elo scores those votes produce. This is also the first time a film company's in-house model has appeared on the leaderboard since it introduced blind evaluation — a debut straight at the summit.
Beyond the scores, what deserves more attention is the identity of this company.
Its business covers film project development, financing, production and global distribution, with both models and platforms serving actual film and series production. For those in the industry, this is a rare sample: how generative AI enters the production process of a complete work from single-shot creation.
A film company has entered the finals of the video model arena
Utopai is headquartered in Mountain View, California, with co-founders Cecilia Shen and Jie Yang serving as CEO and Chief Craft Officer respectively. Cecilia Shen previously worked on data science and AI safety-related approaches at Google X.
Utopai's origins trace back to Cybever, founded in 2022.
Early R&D centred on procedural content generation and high-fidelity 3D environments, laying groundwork for game and film previsualisation, then gradually extending — learning as it went — into spatial reasoning, shot scheduling, and generating cinematic-grade video from 3D previsualisation.
In 2025 Cybever was renamed Utopai Studios, shifting to building a film development, production and distribution system around original works and co-productions. The in-house craft was also brought into real production, used to support the development of characters, scenarios and shots.
According to Forbes, Utopai is valued at around 1 billion dollars. Variety, in its published collaboration feature, describes it as 'the world's largest independent AI-native film company' and introduces three films and two series planned for 2027.
The practical needs of film production posed the questions driving Utopai's model R&D.
One has to remember that with video generation tools, users can repeatedly 'pull cards' to reroll, but feature-length production is a different game: hundreds or even thousands of shots must be strung into one continuous narrative, and characters, environments and props must hold up to the camera being pushed around them.
The production process also involves extensive revision.
A director may approve a character's appearance but want the lighting adjusted; may keep a scene layout but need the character's action changed. Every revision must carry forward the creative choices already settled, so new footage can be used alongside other shots.
Generation quality, cross-shot consistency and modifiability have therefore become the three mountains every feature-length production must climb.
Utopai X's R&D also revolves around these production requirements. Artificial Analysis disclosed that the model is post-trained on MiniMax H3, and this ranking surpassed the original MiniMax H3 as well as MiniMax H3 Max, which is also built on it.
According to the latest published information, Utopai X ranks first in the two capability categories of audio sync and physical performance, and second in lighting and materials and camera control.
Each of these capabilities maps to a hard on-set requirement: sound must match the on-screen action; light passing through glass, water or metal must change convincingly; and when the camera moves, the spatial relationship between character and environment must not break.
Against challenges such as reflections, caustics, spatial understanding and advanced consistency, Utopai X has also been optimised point by point, aiming to improve footage usability so filmmakers can adjust shots around their intent and reduce the work of re-selecting footage after repeated generation.
From Utopai X to PAI, bringing generation capability into the production workflow
What models are able to generate still has to be converted through a production system into capabilities filmmakers can use.
Utopai built the PAI production intelligence platform for this and deeply integrated Utopai X into it. The two handle different work: Utopai X provides video generation capability, while PAI organises scripts, characters, scenarios, shots and creative feedback so the generation process can reuse the project's existing information.
In PAI, a script can be broken down into scenarios and shots, with characters, locations, props and visual references linked to the project throughout. When creators advance to a new shot, they can directly call on confirmed settings instead of retelling the entire story world from scratch every time they generate.
The AI agents within the platform enter as 'production assistants' and are genuine helps: they plan shots, try out creative ideas and produce multiple versions based on production context. Creators, meanwhile, can comparison-shop, roll back to older versions and then decide what to make, keep and change.
Selected footage then enters the editing timeline and connects with professional post-production tools such as Premiere and DaVinci Resolve, forming a coherent workflow from creative development and footage generation through to final delivery.
Throughout, the final say on story, visual style, performance and the shipping version always stays with the filmmakers. Chief Science Officer Zijian He's emphasis that 'craft should expand the filmmaker's canvas' is embodied in this division of labour, giving creators more room to experiment while keeping control of results.
R&D and production run under one roof in parallel, and proximity naturally turned into a continuously iterating flywheel.
The problems artists hit on set become R&D challenges directly — for example, how the same character stays consistent across scenarios, how approved compositions survive after motion changes, how multiple versions enter the edit for the creative team to compare.
The data and expert feedback accumulated in real production are used to refine models, agents and workflows, and the new capabilities that come out of that go back into the current project and roll like a snowball into subsequent works.
Compared with testing only isolated clips, professional production can give more complete feedback. Beyond whether a shot is beautiful, filmmakers also judge whether it connects with adjacent shots, whether it can keep being modified, and whether it meets final delivery requirements.
Chinese IP and global production have a new way of collaborating
Beyond the leaderboard, Utopai's production system is entering stories of different genres and cultural backgrounds, undergoing a tougher test than a single video clip.
The animated feature 'The Most Serious Fart', planned for 2027, is adapted from Mike Bender's 2023 children's picture book of the same name. The story follows an unusually serious fart named Siegfried, who wants to change how the world sees him, and along his journey learns to accept himself and find belonging.
A seemingly absurd premise carries themes of identity, self-acceptance and breaking down prejudice. Utopai hopes to start from distinctive characters and story worlds to develop entertainment IP for global audiences that can be extended over the long term, making a film the starting point of ongoing creation.
The creative team lineup also reflects this development thinking.
Original author Mike Bender will serve as writer and director. He previously worked on the comedy Not Another Teen Movie and is a co-creator of Awkward Family Photos. Russell Hollander serves as producer, and original illustrator Chuck Dillon becomes Creative Art Director, carrying the picture book's character temperament and visual identity into the film.
Joe Stillman, who wrote Shrek and received an Academy Award nomination for Best Adapted Screenplay for it, serves as co-executive producer; Andy Brooks, who worked on How to Train Your Dragon and similar titles, joins on story and storyboard; John Lee, formerly art director at Pixar, is responsible for production design. Casting director Scott Muller, who worked on The Simpsons and Futurama, handles voice casting. The key elements of story, visuals and sound are all handled by
According to Utopai, human animators and illustrators lead the creative work with AI assisting production iteration, while character emotion and story pacing remain in the hands of the creative team.
The family film 'Half Moon' in the same slate involves partners including Germany's In Good Company and Korea's Paper Barn Studios, with PAI providing compatibility for some visual elements. The historical epic 'Cortés' faces a different production challenge — bringing a vast historical scenario into a film that can be completed and released.
The story of 'Cortés' is set in 1519, revolving around Hernán Cortés's expedition to the Aztec Empire and the conflict between Spanish conquerors and Indigenous Americans. Oscar-nominated screenwriter Nicholas Kazan has said he has worked for more than thirty years to tell this story, with the grand scale and high cost always obstacles to progress.
The difficulty of the subject also comes from history itself.
The arrival of the Spanish, the war and the fall of Tenochtitlan involve complex and contested historical accounts. Kazan emphasises that the film hopes to acknowledge the existence of different positions and perspectives, using one concrete story to show two cultures meeting and clashing, and avoiding treating one side's account as the whole of history. This creative stance of accommodating multiple perspectives also means the team must handle historical figures and conflicts carefully — understanding the situations of different groups is as important as reconstructing ancient cities and battle scenes.
Putting the ancient city of Tenochtitlan, the war and imperial change on screen means complex scenario construction, crowd staging and visual production. Utopai plans to use AI production tools to ease these scale and cost constraints, giving new production conditions to a story long unable to reach the screen.
The film is directed by Paul Mignot and produced by Marco Weber, based in Paris, shot in Spain, targeting a 2027 release. The screenwriter's thirty-plus years of persistence and Utopai's investment in production craft are beginning to point toward one outcome: making the story a film audiences can watch.
Utopai's collaborations in Asia are likewise tailored to each work, selecting suitable craft according to the work's narrative needs.
Utopai and Huace Film & TV are jointly developing and producing the animated series 'Journey to the West: The Lost 500 Years', using PAI for development and production; its collaboration with Japan's DeNA covers visual concepts, characters, environments and scenario iteration, with creative approval still held by DeNA's artists.
The Korean business is built on the creative team of Alquimista Media, with APAC head Hyun Park bringing experience from Studio Dragon, Warner Bros. Korea and DramaFever.
AI can help creators test shots faster, but characters, cultural context and narrative trade-offs still need people familiar with the work to sign off.
From Hollywood animation to Chinese classic IP, what I see in these collaborations is not a story of AI upending the film industry. These attempts let production craft adapt to local aesthetics and narrative, blending local creativity with cross-market production and distribution experience — borrowing the boat to set sail.
This may well be a healthier collaboration model for the AI and film industries, and the way these works are distributed is changing too.
'Space Nation', made with partners such as Ex Machina Studios, is a long-form science fiction series planned at eight episodes, revolving around humanity facing an alien threat. For PAI, a series demands far more than generating a few space shots — characters, environments, visual style and narrative continuity must hold up across an entire season.
From worldview and character development to visual construction of different environments and maintaining coherence between episodes, long-form science fiction repeatedly tests whether production tools can be used continuously in complex projects. A series IP that can develop over the long term also needs later stories to reuse existing settings, rather than rebuilding the entire world for every production.
'Space Nation' also became the first AI series to reach agreement with traditional rights buyers, with buyers such as Brazil's Globoplay and German television channels purchasing distribution rights at non-AI market prices.
According to Forbes, with final works delivered roughly as contracted, the potential combined pre-sale revenue of 'Space Nation' and 'Cortés' could reach up to about 110 million dollars.
Beyond the rights deals, advancing the two projects is far less simple than one might imagine.
News of 'Cortés' and 'Space Nation' emerged back in 2025, and only now are they moving toward their planned 2027 targets. The reason is that their long advance has been precisely during a period when Hollywood is especially sensitive about the boundaries of AI use.
According to Forbes, 'Cortés' plans to combine live-action performance with digital environments — filming actors' performances and then placing them into digital scenarios. Such projects require far fewer people than traditional production of equivalent scale. For producers, this is the condition that lets expensive stories get started; for practitioners, the same set of numbers also means job opportunities may change.
The controversy therefore has a concrete target. After live-action filming ends, how far can an actor's likeness be modified, can it be used in other shots or even later works, and how are licensing and payment calculated — all of this directly determines whether such production methods can be accepted.
The agreement reached after the 2023 strikes by SAG-AFTRA has already established informed consent and compensation requirements for digital doubles, with protection covering both leads and extras, so producers cannot read one filming session as permission to use an actor's likeness indefinitely.
Screenwriters have likewise won clear boundaries. The Writers Guild of America's 2023 agreement states that producers cannot force writers to use AI, nor use AI-generated material to undermine a writer's credit and related rights. For studios bringing AI deeply into production, preserving human creators' decision rights needs to be implemented in script development, footage use and contract arrangements.
Facing industry concerns, Cecilia Shen emphasised in a Forbes interview that Utopai uses non-copyright-infringing training data, is willing to work with industry unions, and keeps creative decisions in human hands. These are Utopai's public responses; it cannot be concluded from them that all disagreements have been resolved.
By 2026, Variety's collaboration feature disclosed that 'Space Nation' has entered production and plans to proceed under the Hollywood union framework. From public projects to planned distribution, Utopai's persistence shows up as willingness to invest time in negotiation and production beyond the craft itself — every step forward means simultaneously breaking through in creation, budget, production and industry rules.
And Utopai is also the first AI film company to join the Hollywood writers' guild and the actors' union, effectively obtaining a formal 'household registration' in traditional film circles. Film festivals, professional creators, industry rules and rights deals have step by step welcomed it through the front door of the industry, and the 'AI-native film company' signboard is beginning to correspond to real production and distribution.
New tools rarely kill old art, but they redraw the threshold for telling stories. Just as photography did not drive painting off the stage — instead giving painters more reason to explore what cameras cannot capture — photography gradually developed into its own artistic language.
AI video today is walking that same old road. Utopai has already secured its entry ticket, placing model R&D inside the responsibilities that film production must bear. Whether these works can keep viewers watching, and whether more stories once shut out by budget can actually land, is the real test of this model and platform.
The big-screen experience cannot be fully measured by Elo scores. Only when more stories that were shelved for lack of production conditions can break ground and reach the screen, and viewers are willing to sit through to the credits, has AI truly walked from demo videos into cinema.