{"id":728132,"date":"2026-10-28T15:59:29","date_gmt":"2026-10-28T19:59:29","guid":{"rendered":"https:\/\/www.rochester.edu\/newscenter\/?p=728132"},"modified":"2026-10-28T16:39:26","modified_gmt":"2026-10-28T20:39:26","slug":"how-to-fix-ai-generated-images-pixel-by-pixel-728132","status":"publish","type":"post","link":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/","title":{"rendered":"Improving AI-generated images, pixel by pixel"},"content":{"rendered":"","protected":false},"excerpt":{"rendered":"<p>Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.<\/p>\n","protected":false},"author":1242,"featured_media":728722,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[116],"tags":[24292,18802,18632,24202,18572],"class_list":["post-728132","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-sci-tech","tag-artificial-intelligence","tag-department-of-computer-science","tag-hajim-school-of-engineering-and-applied-sciences","tag-jiebo-luo","tag-research-finding"],"acf":{"external":"","hide_featured_image":true,"related":"show_related","style":[719072,715012,709252],"content_modules":[{"acf_fc_layout":"media_text_new","anchor":"","class":"pt-0","overline":"","title":"","content":"","cta":"","bg":"#ffffff","type":"video","image":"","embed":"","video":728802,"layout":"above","animation":"animation-false"},{"acf_fc_layout":"text","anchor":"","css_class":"pt-0","title":"<h2>Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.<\/h2>","content":"","background_color":"#ffffff","width":"width-medium"},{"acf_fc_layout":"text","anchor":"","css_class":"","title":"","content":"<h3>Why this matters:<\/h3>\r\n<ul>\r\n \t<li>The model <strong>advances AI image generation by allowing users<\/strong> to make precise changes without unintentionally altering other details<strong>.<\/strong><\/li>\r\n<\/ul>\r\n<ul>\r\n \t<li>Users gain greater control by <strong>working directly with individual pixels.<\/strong><\/li>\r\n<\/ul>\r\n<ul>\r\n \t<li>The pixels-space diffusion model <strong>achieves nearly the same performance as latent space models <\/strong>that underpin state-of-the-art image generation tools.<\/li>\r\n<\/ul>\r\n<ul>\r\n \t<li>The pixel-space diffusion model could enable greater creative control, <strong>shaping the future of AI video generation.<\/strong><\/li>\r\n<\/ul>","background_color":"#f4f4f4","width":"width-medium"},{"acf_fc_layout":"text","anchor":"","css_class":"","title":"","content":"If you have ever used an artificial intelligence text-to-image generator like ChatGPT or Google Gemini, you may have encountered a common but maddening problem: the image looks nearly perfect, but when you ask the AI to fix one small detail, it messes up something unrelated.\r\n\r\nThe issue is caused by the way the AI models are built. Most state-of-the-art AI image generators use a diffusion model in latent space, meaning they compress concepts such as a cat or a dog into their essential details and store them so they can be retrieved quickly and efficiently when it\u2019s time to create an image. This is part of what gives the models their seemingly magical ability to generate the images you describe in seconds. But the method has its drawbacks.\r\n\r\n\u201cUsing the latent space helps cut down the computation, but the problem is that you lose controllability because you are not adjusting individual pixels,\u201d says <a href=\"https:\/\/www.cs.rochester.edu\/people\/faculty\/luo_jiebo\/index.html\">Jiebo Luo<\/a>, the Albert Arendt Hopeman Professor of Engineering at the <a href=\"https:\/\/www.rochester.edu\/\">University of Rochester<\/a>. \u201cAs a result, you can\u2019t easily control specific parts of the images you generate.\u201d","background_color":"#ffffff","width":"width-medium"},{"acf_fc_layout":"media_text_new","anchor":"","class":"","overline":"","title":"<h3>Pix(els) or it didn't happen<\/h3>","content":"<strong>Developed by researchers at URochester and NVIDIA, PixelDiT2 efficiently implements pixel space diffusion to achieve <em>almost<\/em> the same performance as the state-of-the-art models\u2014but provides controllability over individual pixels.<\/strong>","cta":"","bg":"#00205b","type":"video","image":"","embed":"","video":728302,"layout":"above","animation":"animation-false"},{"acf_fc_layout":"text","anchor":"","css_class":"","title":"","content":"<h3><strong>Reimagining image generation at the pixel level<\/strong><\/h3>\r\nLuo\u2019s lab and collaborators at the AI infrastructure company <a href=\"https:\/\/www.nvidia.com\/en-us\/\">NVIDIA<\/a> were one of the first teams to develop a model that trades the latent space for the pixel space, dubbed <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2511.20645\">PixelDiT<\/a>. They presented their first iteration at the <a href=\"https:\/\/cvpr.thecvf.com\/Conferences\/2026\">2026 IEEE\/CVF Conference on Computer Vision and Pattern Recognition<\/a> (CVPR), where it was <a href=\"https:\/\/blogs.nvidia.com\/blog\/cvpr-research-grasping-driving-agent-training\/\">nominated as a best paper finalist<\/a>, given to just 15 of over 4,000 accepted papers at the conference. The model allows users to zero in on certain regions of an image, like fixing just the pixels used to illustrate a pair of glasses on a person\u2019s face or adjusting the color of a car.\r\n\r\nThe team has since created an even more powerful model, PixelDiT2, that implements pixel space diffusion in an efficient way to achieve almost the same performance as the state-of-the-art latent space models but providing controllability over individual pixels. The team will present a <a href=\"https:\/\/doi.org\/10.48550\/arXiv.2609.24919\">paper<\/a> showcasing PixelDiT2 at the <a href=\"https:\/\/neurips.cc\/Conferences\/2026\">Neural Information Processing Systems conference<\/a> (NeurIPS) this December.\r\n\r\nLuo and his collaborators, who include <a href=\"https:\/\/www.cs.rochester.edu\/graduate\/phd-program.html\">computer science<\/a> PhD student and NVIDIA intern Yongsheng Yu as well as alumnus Wei Xiong \u201922 PhD (computer science), a senior researcher at NVIDIA, now have turned their attention to an even more difficult problem: video generation.\r\n\r\n\u201cVideo is a big challenge because you have to worry about smooth transitions from frame to frame; otherwise it looks funny to the human eye,\u201d says Luo. \u201cYou also have a lot more data to work with.\u201d\r\n\r\nBut Luo says the progress the team has made with PixelDiT2 leaves them in a strong position to pursue AI video generation in the pixel space, which would provide users powerful new levels of control over the videos they generate.","background_color":"#ffffff","width":"width-medium"}]},"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.6 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Improving AI-generated images, pixel by pixel<\/title>\n<meta name=\"description\" content=\"Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Improving AI-generated images, pixel by pixel\" \/>\n<meta property=\"og:description\" content=\"Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/\" \/>\n<meta property=\"og:site_name\" content=\"News Center\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-28T19:59:29+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-10-28T20:39:26+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.rochester.edu\/newscenter\/wp-content\/uploads\/2026\/10\/fea-Luo-NVIDIA-ai-generated-images-1200x630.gif\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"630\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/gif\" \/>\n<meta name=\"author\" content=\"Luke Auburn\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Luke Auburn\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"1 minute\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/\"},\"author\":{\"name\":\"Luke Auburn\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/#\\\/schema\\\/person\\\/e928dc2863b53a89ece6d40c7992a4e1\"},\"headline\":\"Improving AI-generated images, pixel by pixel\",\"datePublished\":\"2026-10-28T19:59:29+00:00\",\"dateModified\":\"2026-10-28T20:39:26+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/\"},\"wordCount\":6,\"image\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/fea-Luo-NVIDIA-ai-generated-images.gif\",\"keywords\":[\"artificial intelligence\",\"Department of Computer Science\",\"Hajim School of Engineering and Applied Sciences\",\"Jiebo Luo\",\"research finding\"],\"articleSection\":[\"Science &amp; Technology\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/\",\"url\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/\",\"name\":\"Improving AI-generated images, pixel by pixel\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/fea-Luo-NVIDIA-ai-generated-images.gif\",\"datePublished\":\"2026-10-28T19:59:29+00:00\",\"dateModified\":\"2026-10-28T20:39:26+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/#\\\/schema\\\/person\\\/e928dc2863b53a89ece6d40c7992a4e1\"},\"description\":\"Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/fea-Luo-NVIDIA-ai-generated-images.gif\",\"contentUrl\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/fea-Luo-NVIDIA-ai-generated-images.gif\",\"width\":2000,\"height\":1200,\"caption\":\"Gif showing a grid of six boxes turn from fully pixelated to clear images.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Improving AI-generated images, pixel by pixel\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/#website\",\"url\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/\",\"name\":\"News Center\",\"description\":\"University of Rochester\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/#\\\/schema\\\/person\\\/e928dc2863b53a89ece6d40c7992a4e1\",\"name\":\"Luke Auburn\",\"url\":\"https:\\\/\\\/www.rochester.edu\\\/newscenter\\\/author\\\/lauburn\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Improving AI-generated images, pixel by pixel","description":"Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/","og_locale":"en_US","og_type":"article","og_title":"Improving AI-generated images, pixel by pixel","og_description":"Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.","og_url":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/","og_site_name":"News Center","article_published_time":"2026-10-28T19:59:29+00:00","article_modified_time":"2026-10-28T20:39:26+00:00","og_image":[{"width":1200,"height":630,"url":"https:\/\/www.rochester.edu\/newscenter\/wp-content\/uploads\/2026\/10\/fea-Luo-NVIDIA-ai-generated-images-1200x630.gif","type":"image\/gif"}],"author":"Luke Auburn","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Luke Auburn","Est. reading time":"1 minute"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#article","isPartOf":{"@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/"},"author":{"name":"Luke Auburn","@id":"https:\/\/www.rochester.edu\/newscenter\/#\/schema\/person\/e928dc2863b53a89ece6d40c7992a4e1"},"headline":"Improving AI-generated images, pixel by pixel","datePublished":"2026-10-28T19:59:29+00:00","dateModified":"2026-10-28T20:39:26+00:00","mainEntityOfPage":{"@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/"},"wordCount":6,"image":{"@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#primaryimage"},"thumbnailUrl":"https:\/\/www.rochester.edu\/newscenter\/wp-content\/uploads\/2026\/10\/fea-Luo-NVIDIA-ai-generated-images.gif","keywords":["artificial intelligence","Department of Computer Science","Hajim School of Engineering and Applied Sciences","Jiebo Luo","research finding"],"articleSection":["Science &amp; Technology"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/","url":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/","name":"Improving AI-generated images, pixel by pixel","isPartOf":{"@id":"https:\/\/www.rochester.edu\/newscenter\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#primaryimage"},"image":{"@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#primaryimage"},"thumbnailUrl":"https:\/\/www.rochester.edu\/newscenter\/wp-content\/uploads\/2026\/10\/fea-Luo-NVIDIA-ai-generated-images.gif","datePublished":"2026-10-28T19:59:29+00:00","dateModified":"2026-10-28T20:39:26+00:00","author":{"@id":"https:\/\/www.rochester.edu\/newscenter\/#\/schema\/person\/e928dc2863b53a89ece6d40c7992a4e1"},"description":"Researchers from URochester and NVIDIA develop new models that provide pixel-level control over AI image generation.","breadcrumb":{"@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#primaryimage","url":"https:\/\/www.rochester.edu\/newscenter\/wp-content\/uploads\/2026\/10\/fea-Luo-NVIDIA-ai-generated-images.gif","contentUrl":"https:\/\/www.rochester.edu\/newscenter\/wp-content\/uploads\/2026\/10\/fea-Luo-NVIDIA-ai-generated-images.gif","width":2000,"height":1200,"caption":"Gif showing a grid of six boxes turn from fully pixelated to clear images."},{"@type":"BreadcrumbList","@id":"https:\/\/www.rochester.edu\/newscenter\/how-to-fix-ai-generated-images-pixel-by-pixel-728132\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.rochester.edu\/newscenter\/"},{"@type":"ListItem","position":2,"name":"Improving AI-generated images, pixel by pixel"}]},{"@type":"WebSite","@id":"https:\/\/www.rochester.edu\/newscenter\/#website","url":"https:\/\/www.rochester.edu\/newscenter\/","name":"News Center","description":"University of Rochester","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.rochester.edu\/newscenter\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/www.rochester.edu\/newscenter\/#\/schema\/person\/e928dc2863b53a89ece6d40c7992a4e1","name":"Luke Auburn","url":"https:\/\/www.rochester.edu\/newscenter\/author\/lauburn\/"}]}},"_links":{"self":[{"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts\/728132","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/users\/1242"}],"replies":[{"embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/comments?post=728132"}],"version-history":[{"count":17,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts\/728132\/revisions"}],"predecessor-version":[{"id":728812,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts\/728132\/revisions\/728812"}],"acf:post":[{"embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts\/709252"},{"embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts\/715012"},{"embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/posts\/719072"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/media\/728722"}],"wp:attachment":[{"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/media?parent=728132"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/categories?post=728132"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.rochester.edu\/newscenter\/wp-json\/wp\/v2\/tags?post=728132"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}