TIDE: Trajectory-Informed Skill Distillation for Efficient Text-to-Video Prompt Optimization

arXiv:2610.15098v1 Announce Type: new Abstract: Text-to-video (T2V) models are highly sensitive to prompt specificity, while iterative prompt optimization typically improves generation quality only through repeated video generation and vision-language model evaluation at inference time. We propose TIDE, a framework that distills this costly search process into an interpretable and reusable prompt-rewriting Skill. TIDE formulates prompt rewriting as a set of source-conditioned rule modules, each targeting a distinct semantic dimension. Once distilled, the frozen Skill activates the relevant rules and generates both positive and negative prompts in a single LLM call, followed by only one T2V generation, without test-time visual search or backbone fine-tuning. Evolved using only 100 prompts over 14 hours, TIDE achieves prompt-optimization performance comparable to computationally expensive search-based methods. Moreover, when transferred to previously unseen video-generation models, TIDE outperforms existing approaches specifically designed or trained for those models. Code is available at https://github.com/xiningin/TIDE_T2V.

Related

• Spatially-Grounded Text-to-Video Generation via Inference-Time Gradient-Free Optimization
• PEEK: Picking Essential frames via Efficient Knowledge distillation
• A Decomposable Probe for Few-Step Diffusion Models: Prompt, Latent, and Score Selectivity across Backbone Families and Distillation Paradigms
• Guided Trajectory Optimization with Sparse Scaling for Test-Time Diffusion
• TC-Pade: Trajectory-Consistent Pade Approximation for Diffusion Acceleration

Source: arXiv cs.CV | 2026-10-14