--- title: Pushing the Boundaries with Real-Time Video Inpainting description: A team built a real-time video inpainting system capable of erasing objects from live video streams at over 25fps. They rejected diffusion models due to... canonical: https://daily.dev/posts/pushing-the-boundaries-with-real-time-video-inpainting-jprph9l7x twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Pushing the Boundaries with Real-Time Video Inpainting | daily.dev og:description: A team built a real-time video inpainting system capable of erasing objects from live video streams at over 25fps. They rejected diffusion models due to... og:url: https://daily.dev/posts/pushing-the-boundaries-with-real-time-video-inpainting-jprph9l7x og:image: https://api.daily.dev/og/posts/JprPH9L7x.png og:image:alt: Pushing the Boundaries with Real-Time Video Inpainting og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Pushing the Boundaries with Real-Time Video Inpainting **[Software Mansion](https://daily.dev/sources/swmansion)** · 5 min read · 0 upvotes · 0 comments ## Summary A team built a real-time video inpainting system capable of erasing objects from live video streams at over 25fps. They rejected diffusion models due to computational cost and instead used a lightweight spatial-temporal transformer architecture based on DSTT, enhanced with SWIN transformers and Conditional Positional Encodings. Object masks are generated using RTMDet, and inference is accelerated with torch.compile for ~20% speedup. To compensate for the low operating resolution (480x288), full-resolution frames are cached client-side and only masked regions are replaced with inpainted content, significantly improving perceived quality. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [D4RT: Unified, Fast 4D Scene Reconstruction & Tracking](https://daily.dev/posts/d4rt-unified-fast-4d-scene-reconstruction-tracking-q8nmhkzzd) · DeepMind · 0 upvotes · 0 comments - [\[2511.19936\] Image Diffusion Models Exhibit Emergent Temporal Propagation in Videos](https://daily.dev/posts/2511-19936-image-diffusion-models-exhibit-emergent-temporal-propagation-in-videos-yfrd8jmco) · Hacker News · 0 upvotes · 0 comments - [Accelerating Spatio-Temporal Attention for Video Diffusion on TPUs](https://daily.dev/posts/accelerating-spatio-temporal-attention-for-video-diffusion-on-tpus-by8crmpum) · Google Developers · 0 upvotes · 0 comments --- Tags: [#deep-learning](https://daily.dev/tags/deep-learning), [#computer-vision](https://daily.dev/tags/computer-vision) [View this post on daily.dev](https://daily.dev/posts/pushing-the-boundaries-with-real-time-video-inpainting-jprph9l7x) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Pushing the Boundaries with Real-Time Video Inpainting","url":"https://daily.dev/posts/pushing-the-boundaries-with-real-time-video-inpainting-jprph9l7x","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/pushing-the-boundaries-with-real-time-video-inpainting-jprph9l7x"},"datePublished":"2026-08-02T07:15:57.692Z","dateModified":"2026-08-02T07:41:51.287Z","description":"A team built a real-time video inpainting system capable of erasing objects from live video streams at over 25fps. They rejected diffusion models due to...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/ccdbdd7ef21ed57380f2331c50f421fb?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/ccdbdd7ef21ed57380f2331c50f421fb?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Software Mansion","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Software Mansion","logo":"https://media.daily.dev/image/upload/s--sOx0vaVZ--/f_auto,q_auto/v1785654908/logos/swmansion?_a=BAMAMicg0","url":"https://daily.dev/sources/swmansion"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/pushing-the-boundaries-with-real-time-video-inpainting-jprph9l7x","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"deep-learning,computer-vision","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Software Mansion","item":"https://daily.dev/sources/swmansion"},{"@type":"ListItem","position":3,"name":"Pushing the Boundaries with Real-Time Video Inpainting"}]} ```