{"id":7161,"date":"2026-01-04T16:38:05","date_gmt":"2026-01-04T08:38:05","guid":{"rendered":"https:\/\/infernews.com\/?p=7161"},"modified":"2026-01-04T16:38:08","modified_gmt":"2026-01-04T08:38:08","slug":"proedit%ef%bc%9a%e9%96%8b%e6%ba%90%e5%9c%96%e7%89%87%e5%8f%8a%e5%bd%b1%e7%89%87%e7%b7%a8%e8%bc%af","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/proedit%ef%bc%9a%e9%96%8b%e6%ba%90%e5%9c%96%e7%89%87%e5%8f%8a%e5%bd%b1%e7%89%87%e7%b7%a8%e8%bc%af\/","title":{"rendered":"ProEdit\uff1a\u958b\u6e90\u5716\u7247\u53ca\u5f71\u7247\u7de8\u8f2f"},"content":{"rendered":"\n<p><a href=\"https:\/\/isee-laboratory.github.io\/ProEdit\/\" target=\"_blank\" rel=\"noreferrer noopener\">ProEdit<\/a> \u900f\u904e\u00a0<strong>KV-mix<\/strong>\u00a0\u5728\u6ce8\u610f\u529b\u5c64\u878d\u5408\u6e90\/\u76ee\u6a19\u7279\u5fb5\uff0c\u53ca\u00a0<strong>Latents-Shift<\/strong>\u00a0\u64fe\u52d5\u6f5b\u5728\u7a7a\u9593\uff0c\u5be6\u73fe\u9ad8\u4fdd\u771f\u7de8\u8f2f\u3002 \u652f\u63f4 FLUX\u3001HunyuanVideo \u7b49\u6a21\u578b\uff0c\u540c\u6642\u4ea6\u6574\u5408 Qwen3-8B \u89e3\u6790\u81ea\u7136\u8a9e\u8a00\u6307\u4ee4\u3002<\/p>\n\n\n\n<p><a href=\"https:\/\/isee-laboratory.github.io\/ProEdit\/\" target=\"_blank\" rel=\"noreferrer noopener\">ProEdit<\/a> \u89e3\u6c7a\u50b3\u7d71\u53cd\u8f49\u7de8\u8f2f\u904e\u5ea6\u4f9d\u8cf4\u6e90\u5716\u7684\u554f\u984c\uff0c\u80fd\u6e96\u78ba\u8b8a\u63db\u4e3b\u9ad4\u5c6c\u6027\u5982\u59ff\u614b\u3001\u6578\u91cf\u3001\u984f\u8272\uff0c\u540c\u6642\u4fdd\u6301\u80cc\u666f\u4e00\u81f4\u3002 \u9069\u7528\u65bc\u5716\u50cf\u66ff\u63db\uff08\u5982\u8001\u864e\u8b8a\u8c93\u3001\u896f\u886b\u8b8a\u6bdb\u8863\uff09\u8207\u5f71\u7247\u52d5\u614b\u7de8\u8f2f\uff08\u5982\u7d05\u8eca\u8b8a\u9ed1\u8eca\u3001\u9e7f\u8b8a\u725b\uff09\u3002\u9069\u5408 AI \u5167\u5bb9\u5275\u4f5c\u8005\u3001\u5f71\u7247\u5f8c\u88fd\uff0cplug-and-play \u76f8\u5bb9 RF-Solver \u7b49\u5de5\u5177\uff0c\u5728\u591a\u9805\u57fa\u6e96\u6e2c\u8a66\u9054 SOTA \u6548\u80fd\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=\"ProEdit: Inversion-based Editing From Prompts Done Right\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_eDF3x-gyNCA\" 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%2FeDF3x-gyNCA%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/eDF3x-gyNCA\" \/><meta itemprop=\"duration\" content=\"PT1M14S\" \/><meta itemprop=\"uploadDate\" content=\"2025-12-29T05:30:04Z\" \/><\/div><div id=\"lyte_eDF3x-gyNCA\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FeDF3x-gyNCA%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">ProEdit: Inversion-based Editing From Prompts Done Right<\/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\/eDF3x-gyNCA\" 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%2FeDF3x-gyNCA%2F0.jpg\" alt=\"ProEdit: Inversion-based Editing From Prompts Done Right\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"Inversion-based visual editing provides an effective and training-free way to edit an image or a video based on user instructions. Existing methods typically inject source image information during the sampling process to maintain editing consistency. However, this sampling strategy overly relies on source information, which negatively affects the edits in the target image (e.g., failing to change the subject&#039;s atributes like pose, number, or color as instructed). In this work, we propose ProEdit to address this issue both in the attention and the latent aspects. In the attention aspect, we introduce KV-mix, which mixes KV features of the source and the target in the edited region, mitigating the influence of the source image on the editing region while maintaining background consistency. In the latent aspect, we propose Latents-Shift, which perturbs the edited region of the source latent, eliminating the influence of the inverted latent on the sampling. Extensive experiments on several image and video editing benchmarks demonstrate that our method achieves SOTA performance. In addition, our design is plug-and-play, which can be seamlessly integrated into existing inversion and editing methods, such as RF-Solver, FireFlow and UniEdit.\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>ProEdit \u900f\u904e\u00a0KV-mix\u00a0\u5728\u6ce8\u610f\u529b\u5c64\u878d\u5408\u6e90\/\u76ee\u6a19\u7279\u5fb5\uff0c\u53ca\u00a0Latents-Shift\u00a0\u64fe\u52d5\u6f5b\u5728\u7a7a\u9593\uff0c\u5be6\u73fe\u9ad8\u4fdd\u771f\u7de8\u8f2f\u3002 \u652f\u63f4 FLUX\u3001HunyuanVideo \u7b49\u6a21\u578b\uff0c\u540c\u6642\u4ea6\u6574\u5408 Qwen3-8B \u89e3\u6790\u81ea\u7136\u8a9e\u8a00\u6307\u4ee4\u3002 ProEdit \u89e3\u6c7a\u50b3\u7d71\u53cd\u8f49\u7de8\u8f2f\u904e\u5ea6\u4f9d\u8cf4\u6e90\u5716\u7684\u554f\u984c\uff0c\u80fd\u6e96\u78ba\u8b8a\u63db\u4e3b\u9ad4\u5c6c\u6027\u5982\u59ff\u614b\u3001\u6578\u91cf\u3001\u984f\u8272\uff0c\u540c\u6642\u4fdd\u6301\u80cc\u666f\u4e00\u81f4\u3002 \u9069\u7528\u65bc\u5716\u50cf\u66ff\u63db\uff08\u5982\u8001\u864e\u8b8a\u8c93\u3001\u896f\u886b\u8b8a\u6bdb\u8863\uff09\u8207\u5f71\u7247\u52d5\u614b\u7de8\u8f2f\uff08\u5982\u7d05\u8eca\u8b8a\u9ed1\u8eca\u3001\u9e7f\u8b8a\u725b\uff09\u3002\u9069\u5408 AI \u5167\u5bb9\u5275\u4f5c\u8005\u3001\u5f71\u7247\u5f8c\u88fd\uff0cplug-and-play \u76f8\u5bb9 RF-Solver \u7b49\u5de5\u5177\uff0c\u5728\u591a\u9805\u57fa\u6e96\u6e2c\u8a66\u9054 SOTA \u6548\u80fd\u3002<\/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":[169,164,157,120,141],"tags":[],"class_list":["post-7161","post","type-post","status-publish","format-standard","hentry","category-169","category-164","category-157","category-120","category-141"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/7161","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=7161"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/7161\/revisions"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=7161"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=7161"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=7161"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}