{"id":9021,"date":"2026-06-10T23:47:41","date_gmt":"2026-06-10T15:47:41","guid":{"rendered":"https:\/\/infernews.com\/blog\/bernini-is-a-unified-framework-for-video-generation-and-editing-that-combines-an\/"},"modified":"2026-06-10T23:48:43","modified_gmt":"2026-06-10T15:48:43","slug":"bernini-is-a-unified-framework-for-video-generation-and-editing-that-combines-an","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/bernini-is-a-unified-framework-for-video-generation-and-editing-that-combines-an\/","title":{"rendered":"Bernini\uff1a\u5f71\u7247\u751f\u6210\u8207\u7de8\u8f2f\u7684\u65b0\u8def\u7dda"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/pasted-57223d82e833.jpg\" alt=\"Bernini\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Bernini \u662f\u4e00\u500b<strong>\u5f71\u7247\u751f\u6210\u8207\u7de8\u8f2f\u6846\u67b6<\/strong>\uff0c\u6838\u5fc3\u662f\u628a MLLM-based semantic planner \u8207 DiT-based renderer \u7d44\u5408\u8d77\u4f86\uff0c\u8655\u7406\u4e00\u822c\u5f71\u7247\u64f4\u6563\u6a21\u578b\u5e38\u898b\u7684\u5167\u5bb9\u6f02\u79fb\u3001\u6307\u4ee4\u8ddf\u5f9e\u4e0d\u7a69\u5b9a\uff0c\u4ee5\u53ca\u9577\u7247\u6bb5\u898f\u5283\u9b06\u6563\u7b49\u554f\u984c\u3002\u5f9e\u5b9a\u4f4d\u770b\uff0c\u5b83\u4e0d\u662f\u55ae\u7d14\u518d\u5806\u5927\u6a21\u578b\uff0c\u800c\u662f\u5148\u505a\u8a9e\u610f\u898f\u5283\uff0c\u518d\u4ea4\u7531\u751f\u6210\u5668\u843d\u5be6\u756b\u9762\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u500b\u9805\u76ee\u7684\u95dc\u9375\u60f3\u6cd5\uff0c\u5728\u65bc\u300cLatent Semantic Planning\u300d\uff1a\u5148\u5728\u6f5b\u5728\u7a7a\u9593\u5b89\u6392\u8a9e\u610f\uff0c\u518d\u505a video diffusion\u3002\u5c0d\u975e\u7814\u7a76\u80cc\u666f\u8b80\u8005\u4f86\u8aaa\uff0c\u53ef\u4ee5\u7406\u89e3\u70ba\u5148\u5beb\u5206\u93e1\u8349\u7a3f\uff0c\u518d\u9010\u683c\u756b\u9762\u5316\uff0c\u9019\u6bd4\u76f4\u63a5\u7531\u6587\u5b57\u4e00\u6b65\u5230\u4f4d\u751f\u6210\u5f71\u7247\uff0c\u66f4\u6709\u6a5f\u6703\u4fdd\u6301\u6545\u4e8b\u9023\u8cab\u548c\u7de8\u8f2f\u610f\u5716\u4e00\u81f4\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5982\u679c\u60f3\u8a66\uff0c\u8f03\u5408\u7406\u7684\u5207\u5165\u9ede\u662f\u5f71\u7247\u7de8\u8f2f\u4efb\u52d9\uff0c\u4f8b\u5982\u98a8\u683c\u8f49\u63db\u3001\u5b57\u5e55\u6216\u6c34\u5370\u79fb\u9664\u3001\u5c40\u90e8\u4fee\u6539\uff0c\u518d\u89c0\u5bdf\u8f38\u51fa\u6709\u6c92\u6709\u8ddf\u8db3\u6307\u4ee4\u3002\u5009\u5eab\u5217\u51fa\u7684\u74b0\u5883\u504f\u9ad8\u968e\uff0c\u5efa\u8b70\u6e96\u5099 CUDA 12.4\u3001Python 3.11.2\uff0c\u4ee5\u53ca torch==2.5.1+cu124\u3001diffusers==0.35.2\u3001accelerate==0.34.2\u3001transformers==4.57.3\uff1b\u82e5\u6709 H100\u3001H800\u3001H200 \u53ef\u914d\u5408 FlashAttention-3\uff0c\u5176\u4ed6 CUDA GPU \u5247\u9000\u56de FlashAttention-2 \u6216 PyTorch SDPA\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Bernini \u5728 video editing \u7684\u8868\u73fe\u9032\u5165\u90e8\u5206\u4e3b\u6d41 closed-source commercial models \u7684\u7b2c\u4e00\u68af\u968a\uff0c\u8a55\u5206\u4f86\u81ea\u5176\u81ea\u5efa arena\uff0c\u4ee5\u4eba\u5de5\u76f2\u9078\u3001Bradley-Terry score \u53ca pairwise win-rate matrix \u5f59\u6574\u3002\u9019\u985e\u7d50\u679c\u6709\u53c3\u8003\u50f9\u503c\uff0c\u4f46\u66ab\u6642\u4e3b\u8981\u53cd\u6620\u7de8\u8f2f\u5834\u666f\uff1b\u82e5\u4f60\u95dc\u5fc3\u66f4\u8907\u96dc\u7684\u4eba\u7269\u751f\u6210\uff0c\u5b98\u65b9\u4e5f\u63d0\u5230 1.3B \u7684 Bernini-R \u5728\u7c21\u55ae\u4efb\u52d9\u63a5\u8fd1 14B \u7248\u672c\uff0c\u9762\u5c0d\u8907\u96dc\u4efb\u52d9\u4ecd\u6709\u5dee\u8ddd\u3002<\/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=\"Bernini: Latent Semantic Planning for Video Diffusion\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_ZE5YgDYkh_g\" 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%2FZE5YgDYkh_g%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/ZE5YgDYkh_g\" \/><meta itemprop=\"duration\" content=\"PT2M29S\" \/><meta itemprop=\"uploadDate\" content=\"2026-06-03T08:03:13Z\" \/><\/div><div id=\"lyte_ZE5YgDYkh_g\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FZE5YgDYkh_g%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">Bernini: Latent Semantic Planning for Video Diffusion<\/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\/ZE5YgDYkh_g\" rel=\"nofollow noopener\"><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%2FZE5YgDYkh_g%2F0.jpg\" alt=\"Bernini: Latent Semantic Planning for Video Diffusion\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"An open-sourced model for video editing and generation! Project page: https:\/\/bernini-ai.github.io Github: https:\/\/github.com\/bytedance\/Bernini\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n\n<ul class=\"wp-block-list\">\n<li>\u6838\u5fc3\u7d44\u6210\u662f <strong>MLLM-based semantic planner + DiT-based renderer<\/strong><\/li>\n\n\n\n<li>\u5df2\u516c\u958b <strong>Bernini-R<\/strong> \u6b0a\u91cd\uff0c\u5305\u542b <strong>1.3B<\/strong> \u7248\u672c<\/li>\n\n\n\n<li>\u9069\u5408\u7814\u7a76\u5f71\u7247\u751f\u6210\u3001\u5f71\u7247\u7de8\u8f2f\u6d41\u7a0b\uff0c\u6216\u60f3\u6bd4\u8f03\u898f\u5283\u5f0f\u751f\u6210\u65b9\u6cd5\u7684\u4eba<\/li>\n\n\n\n<li>\u786c\u4ef6\u9580\u6abb\u504f\u9ad8\uff0cMulti-GPU sequence parallel \u4ea6\u9700\u8981 Open-VeOmni<\/li>\n\n\n\n<li>\u76f8\u95dc\u6a21\u578b\u53ef\u5148\u7559\u610f <strong>Bernini-R-1.3B-Diffusers<\/strong>\uff0c\u4ee5\u53ca\u6587\u4e2d\u63d0\u5230\u7684 <strong>14B<\/strong> \u8b8a\u9ad4<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u6574\u9ad4\u4f86\u770b\uff0cBernini \u6700\u6709\u50f9\u503c\u7684\u5730\u65b9\u4e0d\u662f\u300c\u518d\u4e00\u500b\u5f71\u7247\u6a21\u578b\u300d\uff0c\u800c\u662f\u628a\u898f\u5283\u8207\u6e32\u67d3\u62c6\u958b\u8655\u7406\uff0c\u4ee4\u53ef\u63a7\u6027\u6210\u70ba\u4e3b\u8981\u8ce3\u9ede\u3002\u82e5\u4f60\u60f3\u627e\u53ef\u76f4\u63a5\u5728\u666e\u901a\u96fb\u8166\u8f15\u9b06\u8dd1\u7684\u9805\u76ee\uff0c\u5b83\u672a\u5fc5\u5408\u9069\uff1b\u4f46\u5982\u679c\u4f60\u91cd\u8996\u7814\u7a76\u65b9\u5411\u3001\u7de8\u8f2f\u8cea\u7d20\u8207\u7cfb\u7d71\u8a2d\u8a08\uff0c\u9019\u500b\u9805\u76ee\u76f8\u7576\u503c\u5f97\u7d30\u770b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/bytedance\/Bernini\" rel=\"noopener noreferrer\">https:\/\/github.com\/bytedance\/Bernini<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Bernini \u662f\u7d71\u4e00\u5f0f\u7684\u5f71\u7247\u751f\u6210\u8207\u7de8\u8f2f\u6846\u67b6\uff0c\u76ee\u6a19\u662f\u4ee4\u751f\u6210\u8207\u4fee\u6539\u7247\u6bb5\u66f4\u53ef\u63a7\u3002\u786c\u4ef6\u8981\u6c42\u4e0d\u4f4e\uff0c\u4f46\u5b9a\u4f4d\u76f8\u7576\u9bae\u660e\u3002<\/p>\n","protected":false},"author":8,"featured_media":9020,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,176,157,120,76,149,141,128,197],"tags":[],"class_list":["post-9021","post","type-post","status-publish","format-standard","hentry","category-133","category-176","category-157","category-120","category-76","category-149","category-141","category-128","category-framework"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9021","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\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=9021"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9021\/revisions"}],"predecessor-version":[{"id":9023,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9021\/revisions\/9023"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9020"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9021"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9021"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9021"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}