Research map / Image generation / Autoregressive and latent generation
Latent Swap Joint Diffusion for 2D Long-Form Latent Generation
Yusheng Dai, Chenxi Wang, Chang Li, Chen Wang, Kewei Li, Jun Du, Lei Sun, Jianqing Gao, Ruoyu Wang, Jiefeng Ma
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 |
|---|---|---|---|
| add Reference-Guided Latent Swap to Self-Loop Swap (ablation) | SaFa (DiT) · Long audio 24s (3x10s views) | I-CLAP: 0.94 -> 0.95 | Improved |
| add Reference-Guided Latent Swap to Self-Loop Swap (ablation) | SaFa (DiT) · Long audio 24s (3x10s views) | I-LPIPS: 0.36 -> 0.34 | Improved |
| add Reference-Guided Latent Swap to Self-Loop Swap (ablation) | SaFa (U-Net) · Long audio 24s (3x10s views) | FD: 7.58 -> 7.88 | Hurt |
| replace averaging with Self-Loop Latent Swap (SaFa* vs MD) (ablation) | MultiDiffusion (DiT) · Long audio 24s (3x10s views) | FD: 11.31 -> 6.98 | Improved |
| replace averaging with Self-Loop Latent Swap (SaFa* vs MD) (ablation) | MultiDiffusion (U-Net) · Long audio 24s (3x10s views) | FAD: 7.06 -> 4.14 | Improved |
| use triangular window in MultiDiffusion (MD* vs MD) | MultiDiffusion (DiT) · Long audio 24s (3x10s views) | FD: 11.31 -> 9.79 | Improved |
| extend generation from 24s to 72s | SaFa (DiT) · Long audio 24s/48s/72s | FD: 6.84 -> 6.98 | No clear effect |
| extend generation from 10s to 96s | SD-audio · AudioCaps long-form | FD: 38.23 -> 30.11 | Mixed |
| extend generation from 10s to 27s | Make-An-Audio2 · AudioCaps long-form | FD: 14.37 -> 18.49 | Hurt |
| apply SaFa joint diffusion instead of MultiDiffusion averaging | Make-An-Audio2 (24s) · AudioCaps 24s on Make-An-Audio2 | FD: 65.28 -> 15.36 | Improved |
| add Reference-Guided Swap to Self-Loop Swap (ablation) | SaFa (SD 3.5 panorama) · Panorama 512x3200 | I-StyleL: 2.96 -> 1.36 | Improved |
| add Reference-Guided Swap to Self-Loop Swap (ablation) | SaFa (SD 2.0 panorama) · Panorama 512x3200 | FID: 32.43 -> 34.71 | Hurt |
Closest papers by meaning
- Seeing and Hearing: Open-domain Visual-Audio Generation with Diffusion Latent Aligners (CVPR 2024 · similarity 0.87)
- MMDisCo: Multi-Modal Discriminator-Guided Cooperative Diffusion for Joint Audio and Video Generation (ICLR 2025 · similarity 0.86)
- Improving the Diffusability of Autoencoders (ICML 2025 · similarity 0.85)
- A Versatile Diffusion Transformer with Mixture of Noise Levels for Audiovisual Generation (NeurIPS 2024 · similarity 0.85)
- Fast Timing-Conditioned Latent Audio Diffusion (ICML 2024 · similarity 0.84)
- Spectral Progressive Diffusion for Efficient Image and Video Generation (NeurIPS 2026 · similarity 0.84)
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.
