LUA Adapter Skips Pixel Upscaling for Diffusion, Speeding Up by 3x
rsasaki0109 · x · 2026-08-23
LUA (Latent Upscale Adapter) is a fast latent space upsampling adapter for diffusion models, accepted to ECCV 2026. It performs ×2/×4 super-resolution directly in the latent space before VAE decoding, requiring no additional diffusion steps. The approach reduces decoding and upscaling time by nearly 3× and transfers to SDXL, SD3, and FLUX with brief fine-tuning, achieving image quality close to native high-resolution generation at 2K and 4K.
Related event: LUA Latent Upscale Adapter Open-Sourced, Speeds Up Diffusion Nearly 3x(2 posts)→
More from Multimodal
- AssemblyAI's Universal-3.6 Pro speech-to-text lands on OpenRouter at $0.30/hr — AssemblyAI · 2026-10-16
- FLUX 3 Image Klein lands on fal: box-drawn layouts, multi-reference placement, style transfer — bfl_ai · 2026-10-16
- SGLang ships Day-0 support for FLUX 3 Image Klein: 10 reference images, BF16/FP8/NVFP4 variants — ying11231 · 2026-10-16
- Krea 3 Scores 4.81 on CPI-Bench, Second Only to GPT Image 2.5 — testingcatalog · 2026-10-16
- Higgsfield teases Kling 4.0 rollout with a monkey-driven demo video — azed_ai · 2026-10-16
- Halcyon early-fusion image models match late-fusion with 275x less pretraining compute, open-sourced — davidwromero · 2026-10-16