Chorus II: Cross-Request Sparsity Reuse for Efficient Image-to-Video Generation

2026-06-23

AI Summary

This work addresses the high computational overhead of diffusion-based image-to-video generation, which hinders efficient deployment in high-concurrency scenarios. The authors propose the first cross-request reuse framework that leverages consistent sparse attention masks from historical requests to eliminate online mask prediction. Combined with optional downsampled feature reuse and a lightweight guidance enhancement mechanism, the approach effectively mitigates semantic drift without requiring any additional training. Under default settings, it achieves a 2.16× speedup in inference while preserving generation quality. The core innovation lies in the first-time exploitation of cross-request consistency in sparse attention patterns to enable efficient video generation.

Abstract

Serving diffusion models for image-to-video generation is computationally expensive, posing significant challenges for large-scale deployment. Real I2V workloads often contain similar requests, such as repeated effect templates, related subjects, and recurring shot layouts. Existing cross-request acceleration methods mainly exploit this redundancy through feature reuse. We observe that similar I2V requests also share highly consistent sparse attention patterns, enabling historical sparse masks to serve as request-conditioned priors with almost no online mask-prediction overhead. We propose a cross-request reuse framework centered on \textbf{sparsity reuse}, with \textbf{feature reuse} as an optional extension safeguarded by a lightweight \textbf{guidance enhancement}. Our sparsity reuse is implemented as shared sparse mask reuse, which reuses high-quality sparse masks from similar historical requests to avoid per-request online mask prediction. Optional feature reuse applies downsampled computation to highly redundant spatiotemporal regions, mitigating boundary artifacts while preserving efficiency gains. Guidance enhancement reinforces image/text conditioning after reuse, mitigating semantic drift and condition-adherence issues. Experiments show that default sparsity reuse configuration preserves generation quality with a \textbf{2.16$\times$} speedup.

Problem

• image-to-video generation
• diffusion models
• cross-request redundancy
• computational efficiency
• sparse attention

Innovation

• sparsity reuse
• cross-request acceleration
• diffusion models
• image-to-video generation
• sparse attention