Project Page: <a href=\"https://siang1105.github.io/JanusMesh.github.io/\" rel=\"nofollow\">https://siang1105.github.io/JanusMesh.github.io/</a><br>Code: <a href=\"https://github.com/siang1105/JanusMesh\" rel=\"nofollow\">https://github.com/siang1105/JanusMesh</a></p>\n","updatedAt":"2026-06-19T03:25:45.488Z","author":{"_id":"6459d5da3b6fafd9664807ab","avatarUrl":"/avatars/57430d1bbde3a2fe5586e5fbcafb0e74.svg","fullname":"Yu-Lun Liu","name":"yulunliu","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":9,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5238121151924133},"editors":["yulunliu"],"editorAvatarUrls":["/avatars/57430d1bbde3a2fe5586e5fbcafb0e74.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.20563","authors":[{"_id":"6a34b61d4c5c5e0d69bf1cb5","name":"Siang-Ling Zhang","hidden":false},{"_id":"6a34b61d4c5c5e0d69bf1cb6","name":"Huai-Hsun Cheng","hidden":false},{"_id":"6a34b61d4c5c5e0d69bf1cb7","name":"Tsung-Ju Yang","hidden":false},{"_id":"6a34b61d4c5c5e0d69bf1cb8","name":"Yu-Lun Liu","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/6459d5da3b6fafd9664807ab/zbYKx8KtWfYw1bSocL23_.qt"],"publishedAt":"2026-06-18T00:00:00.000Z","submittedOnDailyAt":"2026-06-19T00:00:00.000Z","title":"JanusMesh: Fast and Zero-Shot 3D Visual Illusion Generation via Cross-Space Denoising","submittedOnDailyBy":{"_id":"6459d5da3b6fafd9664807ab","avatarUrl":"/avatars/57430d1bbde3a2fe5586e5fbcafb0e74.svg","isPro":false,"fullname":"Yu-Lun Liu","user":"yulunliu","type":"user","name":"yulunliu"},"summary":"Creating 3D visual illusions, a single 3D mesh that reveals entirely different semantics from various viewing angles, is a fascinating but tough challenge. Existing optimization-based methods are slow and can produce oversaturated colors. In contrast, naive stitching approaches fail to produce geometrically coherent objects. This results in visible unnatural seams and semantic leaks. In this paper, we present a fast and training-free framework for generating text-driven 3D visual illusions. Our approach decouples the generation into two stages. First, we propose a cross-space dual-branch denoising process. This process dynamically decodes 3D latents into voxel space for CLIP-guided orientation alignment and Signed Distance Field (SDF) blending, which ensures seamless geometric fusion. Second, we introduce a view-conditioned texture synthesis module that projects and aggregates view-specific 2D diffusion priors onto the fused geometry. Extensive experiments demonstrate that our method generates highly realistic, dual-semantic 3D illusions in just 3-5 minutes. It significantly outperforms existing methods in geometric integrity, semantic recognizability, and efficiency. Project page: https://siang1105.github.io/JanusMesh.github.io/","upvotes":10,"discussionId":"6a34b61d4c5c5e0d69bf1cb9","projectPage":"https://siang1105.github.io/JanusMesh.github.io/","githubRepo":"https://github.com/siang1105/JanusMesh","githubRepoAddedBy":"user","ai_summary":"A fast, training-free framework generates text-driven 3D visual illusions by decoupling generation into cross-space dual-branch denoising and view-conditioned texture synthesis for seamless geometric fusion and semantic coherence.","ai_keywords":["cross-space dual-branch denoising process","3D latents","voxel space","CLIP-guided orientation alignment","Signed Distance Field","SDF blending","view-conditioned texture synthesis","2D diffusion priors","geometric fusion","semantic coherence"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":4,"organization":{"_id":"63e39e6499a032b1c950403d","name":"NYCU","fullname":"National Yang Ming Chiao Tung University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/63e39df6c65f975b436bb6b8/WLWf1bSpvrXBYYKEdXbgU.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6459d5da3b6fafd9664807ab","avatarUrl":"/avatars/57430d1bbde3a2fe5586e5fbcafb0e74.svg","isPro":false,"fullname":"Yu-Lun Liu","user":"yulunliu","type":"user"},{"_id":"6672ebc506b6d49dda7598c5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6672ebc506b6d49dda7598c5/9yUeKzZZVtBoy2L-dNPMf.png","isPro":false,"fullname":"Sytwu","user":"Sytwu","type":"user"},{"_id":"68dff70171512dfe66dad095","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/KtzoaKRYpDVZLYvk00VL8.png","isPro":false,"fullname":"李佑軒","user":"johnnyli1220","type":"user"},{"_id":"64cdecee2f1f9578a0e701c8","avatarUrl":"/avatars/95a51dd4e1b7b9366ebcbd6028ad148b.svg","isPro":false,"fullname":"Yi-Ruei Liu","user":"Shigon","type":"user"},{"_id":"68ebb3d0c9dda63b9d7f4a85","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/MFsp2eYyc6KdGDP5m9Y1a.png","isPro":false,"fullname":"曾士珍","user":"sjtseng","type":"user"},{"_id":"6407e5294edf9f5c4fd32228","avatarUrl":"/avatars/8e2d55460e9fe9c426eb552baf4b2cb0.svg","isPro":false,"fullname":"Stoney Kang","user":"sikang99","type":"user"},{"_id":"670753680681f4d0a94ebccf","avatarUrl":"/avatars/1aa6f063bacdb25d36784d0f93bb2224.svg","isPro":true,"fullname":"ChengYou Lu","user":"ChengYou305","type":"user"},{"_id":"6717a0aebc4492cad1a36a65","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/B0u38BtEvw_32bkovDGI2.png","isPro":false,"fullname":"Siang Ling Zhang","user":"Siangling","type":"user"},{"_id":"66bc5e57432c73a183ec3c15","avatarUrl":"/avatars/aad325111bfc9a29a8eda8dc36b15413.svg","isPro":false,"fullname":"Chih Yu Chang","user":"qomolanma","type":"user"},{"_id":"660c483a9d3a84c103fc43ed","avatarUrl":"/avatars/5aa906f402e18fb05c28a6dda3962d1c.svg","isPro":false,"fullname":"Chia-Ming Lee","user":"ming0531","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"63e39e6499a032b1c950403d","name":"NYCU","fullname":"National Yang Ming Chiao Tung University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/63e39df6c65f975b436bb6b8/WLWf1bSpvrXBYYKEdXbgU.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.20563.md","query":{}}">
JanusMesh: Fast and Zero-Shot 3D Visual Illusion Generation via Cross-Space Denoising
Abstract
A fast, training-free framework generates text-driven 3D visual illusions by decoupling generation into cross-space dual-branch denoising and view-conditioned texture synthesis for seamless geometric fusion and semantic coherence.
Creating 3D visual illusions, a single 3D mesh that reveals entirely different semantics from various viewing angles, is a fascinating but tough challenge. Existing optimization-based methods are slow and can produce oversaturated colors. In contrast, naive stitching approaches fail to produce geometrically coherent objects. This results in visible unnatural seams and semantic leaks. In this paper, we present a fast and training-free framework for generating text-driven 3D visual illusions. Our approach decouples the generation into two stages. First, we propose a cross-space dual-branch denoising process. This process dynamically decodes 3D latents into voxel space for CLIP-guided orientation alignment and Signed Distance Field (SDF) blending, which ensures seamless geometric fusion. Second, we introduce a view-conditioned texture synthesis module that projects and aggregates view-specific 2D diffusion priors onto the fused geometry. Extensive experiments demonstrate that our method generates highly realistic, dual-semantic 3D illusions in just 3-5 minutes. It significantly outperforms existing methods in geometric integrity, semantic recognizability, and efficiency. Project page: https://siang1105.github.io/JanusMesh.github.io/
Community
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2606.20563 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.20563 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.20563 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.