{"id":6442,"date":"2025-09-20T00:06:00","date_gmt":"2025-09-19T16:06:00","guid":{"rendered":"https:\/\/infernews.com\/?p=6442"},"modified":"2025-09-20T00:07:48","modified_gmt":"2025-09-19T16:07:48","slug":"nvidia-cosmos-%e9%87%8d%e7%b5%84%e5%85%89%e6%ba%90","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/nvidia-cosmos-%e9%87%8d%e7%b5%84%e5%85%89%e6%ba%90\/","title":{"rendered":"NVidia Cosmos\u00a0\u91cd\u7d44\u5149\u6e90"},"content":{"rendered":"\n<p>\u958b\u6e90 <strong><a href=\"https:\/\/research.nvidia.com\/labs\/toronto-ai\/DiffusionRenderer\/\">Cosmos<\/a><\/strong>&nbsp;DiffusionRenderer \u662f\u4e00\u500b\u8996\u8a0a\u64f4\u6563\u6846\u67b6\uff0c\u7528\u65bc\u9ad8\u54c1\u8cea\u5f71\u50cf\u548c\u8996\u8a0a\u7684\u53bb\u5149\u548c\u91cd\u5149\u3002\u5b83\u662f\u539f\u59cb<br>DiffusionRenderer \u7684\u91cd\u5927\u66f4\u65b0\uff0c\u5728 NVIDIA \u6539\u9032\u7684\u8cc7\u6599\u7ba1\u7406\u6d41\u7a0b\u7684\u652f\u6301\u4e0b\uff0c\u5be6\u73fe\u4e86\u986f\u8457\u66f4\u9ad8\u54c1\u8cea\u7684\u7d50\u679c\u3002<\/p>\n\n\n\n<pre>\n\u6700\u4f4e\u8981\u6c42\nPython 3.10\nNVIDIA GPU \u81f3\u5c11\u914d\u5099 16GB VRAM\uff0c\u5efa\u8b70\u914d\u5099 >=48GB VRAM\nNVIDIA \u9a45\u52d5\u7a0b\u5f0f\u548c CUDA 12.0 \u6216\u66f4\u9ad8\u7248\u672c\n\u81f3\u5c11 70GB \u53ef\u7528\u78c1\u789f\u7a7a\u9593<\/pre>\n\n\n\n<figure class=\"wp-block-video\"><video height=\"398\" style=\"aspect-ratio: 1074 \/ 398;\" width=\"1074\" controls src=\"\/blog\/wp-content\/uploads\/2025\/09\/teaser.mp4\"><\/video><\/figure>\n\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"Relight Any Scene with AI: NVIDIA Research Unveils DiffusionRenderer\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_Q3xhYNbXM9c\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQ3xhYNbXM9c%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/Q3xhYNbXM9c\" \/><meta itemprop=\"duration\" content=\"PT58S\" \/><meta itemprop=\"uploadDate\" content=\"2025-06-16T23:12:23Z\" \/><\/div><div id=\"lyte_Q3xhYNbXM9c\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQ3xhYNbXM9c%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">Relight Any Scene with AI: NVIDIA Research Unveils DiffusionRenderer<\/div><\/div><div class=\"play\"><\/div><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/Q3xhYNbXM9c\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQ3xhYNbXM9c%2F0.jpg\" alt=\"Relight Any Scene with AI: NVIDIA Research Unveils DiffusionRenderer\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"Discover DiffusionRenderer, a groundbreaking neural rendering technique from NVIDIA Research that acts like a virtual light switch for video. It unifies inverse and forward rendering in a single AI-powered engine, enabling realistic, dynamic lighting transitions\u2014from day to night, sunny to cloudy\u2014and outperforms existing methods in physical AI training environments. Paper: https:\/\/research.nvidia.com\/labs\/toronto-ai\/DiffusionRenderer\/ #nvidia #nvidiaresearch #generativeai\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n\n<p>\u6559\u5b78\uff1a<\/p>\n\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"NVIDIA&amp;#039;s new AI Just Made Any Video Relightable (DiffusionRenderer Demo)\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_Qjzc_uORnuE\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQjzc_uORnuE%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/Qjzc_uORnuE\" \/><meta itemprop=\"duration\" content=\"PT37M10S\" \/><meta itemprop=\"uploadDate\" content=\"2025-09-18T19:00:32Z\" \/><\/div><div id=\"lyte_Qjzc_uORnuE\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQjzc_uORnuE%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">NVIDIA&#039;s new AI Just Made Any Video Relightable (DiffusionRenderer Demo)<\/div><\/div><div class=\"play\"><\/div><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/Qjzc_uORnuE\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQjzc_uORnuE%2F0.jpg\" alt=\"NVIDIA&amp;#039;s new AI Just Made Any Video Relightable (DiffusionRenderer Demo)\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"To learn for free on Brilliant, go to https:\/\/brilliant.org\/PixelReconstruct\/ . You\u2019ll also get 20% off an annual premium subscription. What if you could turn any video into golden hour\u2026 after you filmed it? NVIDIA\u2019s new DiffusionRenderer makes that possible. Relighting videos with physics-accurate shadows, reflections, and materials. In this deep dive, I\u2019ll show you: - How inverse rendering + forward rendering power DiffusionRenderer - Why this matters for filmmaking, VFX, and autonomous robotics - Comparison of real world results from images and video input This isn\u2019t just about pretty pictures \u2014 it could change how Hollywood lights scenes, and how companies like Tesla &amp; Waymo train self-driving cars. \ud83d\udd17 Resources &amp; Code: DiffusionRenderer on GitHub: https:\/\/github.com\/nv-tlabs\/cosmos1-diffusion-renderer DiffusionRenderer Project Page: https:\/\/research.nvidia.com\/labs\/toronto-ai\/DiffusionRenderer\/ Follow me: LinkedIn: https:\/\/www.linkedin.com\/in\/jonathanstephens\/ X (Twitter): https:\/\/twitter.com\/jonstephens85 \ud83d\udc49 Subscribe to Pixel Reconstruct for more hands-on breakdowns of the latest in 3D AI, computer vision, and VFX tech. This video was sponsored by Brilliant. Thank you for making this content possible. #nvidia #ai #genai #vfx #nvidiaai\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>","protected":false},"excerpt":{"rendered":"<p>\u958b\u6e90 Cosmos&nbsp;DiffusionRenderer \u662f\u4e00\u500b\u8996\u8a0a\u64f4\u6563\u6846\u67b6\uff0c\u7528\u65bc\u9ad8\u54c1\u8cea\u5f71\u50cf\u548c\u8996\u8a0a\u7684\u53bb\u5149\u548c\u91cd\u5149\u3002\u5b83\u662f\u539f\u59cbDiffusionRenderer \u7684\u91cd\u5927\u66f4\u65b0\uff0c\u5728 NVIDIA \u6539\u9032\u7684\u8cc7\u6599\u7ba1\u7406\u6d41\u7a0b\u7684\u652f\u6301\u4e0b\uff0c\u5be6\u73fe\u4e86\u986f\u8457\u66f4\u9ad8\u54c1\u8cea\u7684\u7d50\u679c\u3002 \u6700\u4f4e\u8981\u6c42 Python 3.10 NVIDIA GPU \u81f3\u5c11\u914d\u5099 16GB VRAM\uff0c\u5efa\u8b70\u914d\u5099 >=48GB VRAM NVIDIA \u9a45\u52d5\u7a0b\u5f0f\u548c CUDA 12.0 \u6216\u66f4\u9ad8\u7248\u672c \u81f3\u5c11 70GB \u53ef\u7528\u78c1\u789f\u7a7a\u9593 \u6559\u5b78\uff1a<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"googlesitekit_rrm_CAowvqSiDA:productID":"","footnotes":""},"categories":[168,120,27,141,133],"tags":[],"class_list":["post-6442","post","type-post","status-publish","format-standard","hentry","category-linux","category-120","category-paper","category-141","category-133"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/6442","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=6442"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/6442\/revisions"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=6442"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=6442"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=6442"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}