即梦AI综合其他
#综合其他
中文翻译
人工智能刚刚彻底摧毁了整个视觉特效预览流程
过去,一个工作室每拍一个镜头要花费40万美元、动用30位专家:
→ 概念艺术:2周
→ 3D粗模与布局:1周
→ 角色建模与绑定:3周
→ 灯光、贴图、最终渲染:2周
→ 合成与调色:1周
这段视频展示了什么:
→ Blender:一个粉色人体模型走过白色方块建筑。没有贴图,没有脸,没有头发。一个一年级学生20分钟就能做出的3D草图。
→ Seedance 2.5:把同样的线框预览变成了一位穿着飘逸丝绸裙的女孩,走在霓虹赛博朋克城市上空,有体积雾、玻璃上的光线反射,还有逐帧追踪原始人体模型动作的头发物理效果。
预览就是最终镜头。没有建模流程。没有贴图师。没有渲染农场。
总时间:不到30分钟。
总成本:一个Seedance订阅。
一个学生。一个Blender场景。一个提示词。输出看起来像是出自一部耗资5000万美元的动漫长片。
在他们删掉之前保存这个
原帖
AI JUST KILLED THE ENTIRE VFX PREVIS PIPELINE
what a studio used to spend $400,000 and 30 specialists on per shot:
→ Concept Art: 2 weeks
→ 3D Blockout & Layout: 1 week
→ Character Modeling & Rigging: 3 weeks
→ Lighting, Texturing, Final Render: 2 weeks
→ Compositing & Color Grade: 1 week
what this clip shows:
→ Blender: a pink mannequin walking through white box buildings. no textures, no face, no hair. a 3D sketch a first-year student could make in 20 minutes.
→ Seedance 2.5: takes that same wireframe preview and turns it into a girl in a flowing silk dress walking above a neon cyberpunk city with volumetric fog, light reflections on glass, and hair physics that track the original mannequin's motion frame by frame.
the previs IS the final shot. no modeling pipeline. no texture artists. no render farm.
total time: under 30 minutes.
total cost: one Seedance subscription.
one student. one Blender scene. one prompt. the output looks like it came off a $50,000,000 anime feature.
save this before they take it down
