NoisEasier: Test-Time Noise Optimization for Text-to-Video Generation
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address problems in text-to-video generation where models have trouble correctly showing specific details like object traits and interactions. They introduce NoisEasier, a method that improves video creation during the model's testing phase without changing the model itself, by carefully adjusting noise based on rewards. This approach helps the videos better match the text prompts, working faster and more reliably than previous methods. Tests show their method improves performance on important aspects like attribute matching and interactions by over 10%. NoisEasier offers a new way to improve video generation that complements existing fine-tuning techniques.
diffusion modelstext-to-video generationcompositional alignmentreward-based fine-tuningtest-time optimizationnoise optimizationlatent spaceattribute bindingobject interactionvideo synthesis
Authors
Yujiang Pu, Yu Kong
Abstract
Diffusion models have recently advanced text-to-video (T2V) generation, yet they still struggle with fine-grained compositional alignment, such as attribute binding, spatial relations, and object interactions. While reward-based fine-tuning improves alignment, it is susceptible to reward hacking and adapts poorly to new prompt distributions. In this work, we propose NoisEasier, a test-time scaling framework that improves T2V generation through differentiable reward-guided noise optimization without modifying the underlying model. By combining efficient short-step generators with a multi-objective reward formulation, NoisEasier enables stable and practical test-time optimization under realistic inference budgets. Our key insight is that jointly optimizing the entire stochastic trajectory accelerates reward convergence and improves compositional alignment over optimizing only the initial latent, with negligible additional computational and time cost. Experiments on VBench and T2V-CompBench demonstrate consistent improvements across multiple backbones, achieving over 10% average gains on challenging dimensions such as attribute binding, object interaction, and numeracy. Overall, NoisEasier serves as both a flexible alternative and a complementary enhancement to reward-based fine-tuning, establishing test-time scaling as an effective paradigm for controllable text-to-video generation.