Search papers, labs, and topics across Lattice.
This paper introduces DUET, a novel framework that combines trajectory-level and distribution-level distillation for two-step video generation, effectively reconciling the quality-diversity trade-off inherent in existing methods. By employing a dual-expert system鈥攁n sCM expert for high-noise diversity and a DMD expert for low-noise quality refinement鈥擠UET achieves significant improvements in both structural diversity and visual quality. The results demonstrate that DUET can elevate the quality of sCM-generated videos to near DMD levels while maintaining approximately double the diversity, with further enhancements in DUET+ through RL-guided expert adaptation.
DUET achieves a groundbreaking balance in video generation, delivering high-quality visuals while preserving twice the diversity of traditional methods.
Diffusion models have enabled high-quality video generation in recent years, but the high cost of iterative sampling hinders their practical deployment. Few-step distillation alleviates this cost, yet exposes a quality--diversity trade-off between its two dominant paradigms: trajectory-level distillation (e.g., sCM) favors diversity, whereas distribution-level distillation (e.g., DMD) favors quality. Targeting extreme two-step video generation, we introduce DUET, which reconciles the two paradigms through a noise-level duet of experts: an sCM expert takes the high-noise step to lay out diverse structure, and a DMD expert takes the low-noise step to refine appearance detail. Since the two experts are trained independently with their native objectives, DUET sidesteps the optimization difficulties of loss-level combinations and delivers quality and diversity jointly rather than trading one for the other. We further identify the relay interface and the high-noise stage as the remaining bottlenecks, and address them with RL-guided expert adaptation, yielding DUET+. With the Wan2.1-T2V-1.3B backbone, DUET lifts the two-step quality of sCM close to the level of DMD while retaining nearly all of its structural diversity---about twice that of DMD---and DUET+ further improves overall quality while preserving this diversity advantage. Together, these results establish noise-level expert specialization as a simple, effective paradigm for reconciling diversity and quality in two-step video generation.
