# Sink-Sparse Hybrid Attention speeds up text-to-image and video diffusion models

> A new train‑free attention method cuts inference time by up to 1.57× on two diffusion transformers while keeping quality.

Oossa · 2026-10-19 · https://oossa.com/en/sink-sparse-hybrid-attention-speeds-up-text-to-image-and-video-diffusion-models

Researchers led by Bingxin Xing introduced Sink‑Sparse Hybrid Attention, a way to replace dense self‑attention in pretrained diffusion transformers without retraining. The technique builds static “sink masks” from a small calibration set and adds a light pooled‑KV branch. On the Wan2.1‑T2V‑1.3B model it speeds up end‑to‑end generation by 1.54×, and on Z‑Image‑Turbo by 1.57×, with quality comparable to the original models.

## The facts

- Paper submitted 16 Oct 2026
- 1.54× speedup on Wan2.1‑T2V‑1.3B, 1.57× on Z‑Image‑Turbo

## Why it matters

Faster inference lets developers generate images or video more cheaply, which can lower costs for apps that rely on text‑to‑media generation.

## Sources & references

1. [Sink-Sparse Hybrid Attention for Efficient Inference in T2I/T2V Diffusion Transformers](https://arxiv.org/abs/2610.22975) – arXiv cs.CV, 2026-10-19

Last updated: 2026-10-19