NVIDIA’s Sol-Attn speeds up video generation inference by 2.1x
Haopeng Li · hf · 2026-07-28
NVIDIA proposes Sol-Attn for faster video generation inference
The paper targets the attention bottleneck in diffusion transformers used for high-fidelity video generation. It introduces Sol-Attn (“Sparsifying online attention”), a training-free method that combines dynamic routing, sparse computation, and approximation correction in a single online-softmax pass.
What it changes
- Uses on-the-fly block thresholding instead of fixed top-k or cumulative-mass routing.
- Avoids materializing a proxy map by reusing proxy scores during routing.
- Approximates the contribution of dropped blocks instead of discarding them outright.
Reported results
- Improves the quality-efficiency trade-off for sparse attention.
- Delivers 2.1× end-to-end speedup for video generation.
- Delivers 2.3× end-to-end speedup for video editing.
- Evaluated across both image and video generation tasks while preserving visual quality.
Related event: NVIDIA Introduces Sol-Attn to Accelerate Video Generation(2 posts)→
More from Multimodal
- Post says AI is replacing ray tracing in real time — jongranskog · 2026-07-28
- Happy Horse shows video editing that preserves motion and micro-expressions — umesh_ai · 2026-07-28
- An AI visual feed idea would turn character dossiers into ComfyUI-generated posts — tovarischsht · 2026-07-28
- An AI-made chocolate meme turns a holiday shoutout into a cozy joke — LudovicCreator · 2026-07-28
- GPT-5.6-powered three.js map builder generates terrain, water, and vegetation — Dimillian · 2026-07-28
- Users ask how to keep characters consistent in LTX2.3 image-to-video workflows — Adventurous-Gold6413 · 2026-07-28