The paper introduces Parallel Decoding Distillation (PDD), a trajectory-based distillation method for diffusion and flow-matching models. PDD predicts multiple denoising steps per network evaluation and supports a variable number of function evaluations (NFE). The authors report state-of-the-art results at 4–8 NFE on LTX-2.3 text-to-video/audio, Wan 14B text-to-video, and Qwen-Image text-to-image. Unlike methods relying heavily on variational score distillation or adversarial losses, PDD is designed to avoid derivative regression with JVPs or finite differences and is reported to improve generated-video diversity.
No heat snapshots are available in the last 24 hours.