AI影视大模型

Search documents
创意、拍摄、后期全流程覆盖 ⻁鲸⽂娱AI影视⼤模型亮相云栖⼤会
Xin Lang Ke Ji· 2025-09-26 05:35
镜⼀介绍,在中期拍摄阶段,⻁鲸⽂娱研发了图⽣3D场景技术模型,可⽀持快速⽣成多机位分镜参考 图。主创若想查看其他视⻆的分镜效果,⽆需重新绘制2D分镜或搭建3D场景,模型能基于单张图⽚重 构3D空间,同步⽣成多视⻆画⾯,还⽀持任意⻆度推拉摇移运镜,帮助主创快速确认空间布局。 新浪科技讯 9月26日午间消息,在杭州举办的云栖⼤会AIGC创新发展论坛上,⻁鲸⽂娱集团AI影视⼯ 业化算法负责⼈镜⼀发表《影视产业中的⼤模型应⽤探索》演讲,⾸次对外介绍⽬前⻁鲸⽂娱在AI影 视⼤模型中的技术进展。 镜⼀表示,"⻁鲸⽂娱与通义万相协同研发的AI影视⼤模型系列,已覆盖前期创意-中期拍摄-后期制作 的全流程,助⼒主创团队推出更优质内容。"据了解,⻁鲸⽂娱的AI影视⼤模型包括运镜模型 CameraWeaver、⽆穿戴AI⾯部动捕、AI表情增强等技术上的突破,能够帮助影视主创团队在创意阶段 快速落地分镜构想、拍摄期间⾼效⽣成3D场景、动漫后期制作⾃动模拟优化微表情。 "尽管⽬前市⾯上已有多个⽂⽣图模型能够相对精细地⽣成图⽚,但在主创更关注的景别、构图视⻆等 ⽅⾯仍⽆法达到要求。"镜⼀说,针对这⼀难题,⻁鲸⽂娱研发的"影视级⽂⽣图"⼤ ...
2025云栖大会:⻁鲸⽂娱AI影视⼤模型已覆盖创意、拍摄、后期全流程
Huan Qiu Wang· 2025-09-25 09:01
【环球网科技报道 记者 李文瑶】"⻁鲸⽂娱与通义万相协同研发的 AI 影视⼤模型系列,已覆盖'前期创意—中期拍摄—后期制作'全流程,助⼒主创团队推 出更优质内容。"9 ⽉ 25 ⽇,在杭州举办的云栖⼤会 AIGC 创新发展论坛上,⻁鲸⽂娱集团 AI影视⼯业化算法负责⼈镜⼀发表《影视产业中的⼤模型应⽤探 索》演讲,⾸次对外介绍⽬前⻁鲸⽂娱在 AI影视⼤模型中的技术进展。 据了解, ⻁鲸⽂娱的这套 AI 影视⼤模型,包括运镜模型 CameraWeaver、⽆穿戴 AI ⾯部动捕技术、AI 表情增强等技术上的突破,能够帮助影视主创团队 在创意阶段快速落地分镜构想、拍摄期间⾼效⽣成 3D 场景、动漫后期制作⾃动模拟优化微表情。 在影视⾏业的内容创意阶段,主创团队通常需要⽤语⾔描述结合⼿绘的⽅式将预想中的场景画⾯呈现出来,往往耗费较多时间。"尽管⽬前市⾯上已有多个 ⽂⽣图模型能够相对精细地⽣成图⽚,但在主创更关注的景别、构图视⻆等⽅⾯仍⽆法达到要求。"镜⼀说,针对这⼀难题,⻁鲸⽂娱研发的"影视级⽂⽣ 图"⼤模型,区别于市⾯通⽤模型追求"视觉发散",在影视级要求的精准控制上做了针对性调试,精准解决了主创"创意画⾯难落地" ...
创意、拍摄、后期全流程覆盖,虎鲸文娱AI影视大模型亮相云栖大会
3 6 Ke· 2025-09-25 08:43
Core Viewpoint - The collaboration between Tiger Whale Entertainment and Tongyi Wanxiang has led to the development of an AI film model series that covers the entire process from pre-creation to post-production, enhancing the quality of content produced by creative teams [1][3]. Group 1: AI Film Model Development - Tiger Whale Entertainment's AI film model series includes innovations such as CameraWeaver for shot planning, non-wearable AI facial motion capture technology, and AI expression enhancement, which streamline the filmmaking process [1][3]. - The "film-level text-to-image" model developed by Tiger Whale Entertainment addresses the industry's pain points by providing precise control over scene composition and perspective, significantly improving the efficiency of pre-creation communication and storyboard development [3][6]. Group 2: Mid-Production Innovations - During the mid-production phase, the company has created a 3D scene generation model that allows for rapid creation of multi-angle storyboard reference images, achieving high precision in 3D space reconstruction within 2 minutes, compared to traditional methods that take weeks [6][7]. - This technology has already been applied in internal projects, achieving top scores in WorldScore during testing, demonstrating its effectiveness in virtual shooting rehearsals [6]. Group 3: Post-Production Enhancements - In post-production, Tiger Whale Entertainment has introduced a non-wearable facial capture and AI refinement solution that accurately restores and completes actors' micro-expressions and lip movements, significantly enhancing the efficiency of animation work [7][10]. - The AI expression generation model automates the process that previously required weeks of manual adjustments by animators, thus improving the quality and speed of micro-expression production [7]. Group 4: Overall Impact on Creators - The company emphasizes that AI is not meant to replace creators but to liberate them from technical burdens, allowing them to focus on storytelling, similar to how calculators assist accountants without replacing them [10].