Research map / Image generation / Autoregressive and latent generation
SphereDiff: Tuning-free 360° Static and Dynamic Panorama Generation via Spherical Latent Representation
Minho Park, Taewoong Kang, Jooyeol Yun, Sungwon Hwang, Jaegul Choo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use spherical latents instead of ERP-latent DynamicScaler (static) | tuning-free MultiDiffusion · 360 static (automatic eval) | Distortion (VLM judge): 2.854 -> 3.238 | Improved |
| use spherical latents instead of ERP-latent DynamicScaler (static) | tuning-free MultiDiffusion · 360 static (automatic eval) | End continuity (VLM judge): 3.985 -> 4.892 | Improved |
| use spherical latents instead of ERP-latent DynamicScaler (live) | tuning-free MultiDiffusion · 360 live (automatic eval) | End continuity (VLM judge): 2.971 -> 4.496 | Improved |
| use spherical latents instead of ERP-latent DynamicScaler (live) | tuning-free MultiDiffusion · 360 live (automatic eval) | Distortion (VLM judge): 1.971 -> 2.579 | Improved |
| SphereDiff vs DynamicScaler on static image quality | tuning-free MultiDiffusion · 360 static (automatic eval) | Image quality (VLM judge): 4.496 -> 4.496 | No clear effect |
| SphereDiff vs 360 LoRA + AnimateDiff on live image quality | 360 live generation · 360 live (automatic eval) | Image quality (VLM judge): 3.179 -> 3.050 | Hurt |
| SphereDiff vs DynamicScaler on live temporal flickering | 360 live generation · 360 live (automatic eval) | Temporal flickering (VBench): 0.9918 -> 0.9941 | Improved |
| SphereDiff vs DynamicScaler on static text alignment | 360 static generation · 360 static (automatic eval) | CLIP-Score: 26.63 -> 28.65 | Improved |
| add dynamic latent sampling instead of nearest sampling (ablation) | SphereDiff · two-view ablation (qualitative) | visual seamlessness | Improved |
| add distortion-aware weighted averaging (ablation) | SphereDiff · two-view ablation (qualitative) | visual clarity/seamlessness | Improved |
Later papers that compared with this method
1 later accepted paper reports SphereDiff as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Linear Fusion MultiDiffusion for Fast Training-Free Spherical Panorama Generation (ECCV 2026 · better on 6 of 7 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SphereDiffusion: Spherical Geometry-Aware Distortion Resilient Diffusion Model (AAAI 2024 · similarity 0.89)
- HORIZON: High-Resolution Semantically Controlled Panorama Synthesis (AAAI 2024 · similarity 0.88)
- Spherical Manifold Guided Diffusion Model for Panoramic Image Generation (CVPR 2025 · similarity 0.87)
- ViewPoint: Panoramic Video Generation with Pretrained Diffusion Models (NeurIPS 2025 · similarity 0.87)
- Spherical-Nested Diffusion Model for Panoramic Image Outpainting (ICML 2025 · similarity 0.87)
- CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
