High-Fidelity Two-Step Image Generation via Teacher-Aligned End-to-End Distillation

2026-06-10

AI Summary

This work addresses the performance limitations of two-step high-fidelity image generation, which stem from the inherent difficulty of the task and model capacity constraints. To bridge this gap, the authors propose a two-step generation framework based on diffusion distillation, incorporating three key innovations: using images synthesized by a teacher model as the target for adversarial training to achieve distribution alignment, assigning dedicated parameters to each denoising step to enhance modeling capacity, and enabling end-to-end training by jointly optimizing final image quality and intermediate-step losses with iterative regularization. Experimental results demonstrate that the proposed method significantly narrows the performance gap between two-step and eight-step generation in both qualitative and quantitative evaluations, closely approaching the fidelity of the teacher model.

Abstract

Few-step diffusion distillation has become increasingly mature for 4-8-step generation, yet pushing further to 2 steps remains challenging. In this work, we introduce Z-Image Turbo++, a high-quality 2-step image generation model distilled from the 8-step Z-Image Turbo teacher. Our method addresses the central bottlenecks of increased task difficulty and limited model capacity in 2-step generation through three simple but effective design choices tailored to this regime. First, we propose Distribution-Aligned Adversarial Learning, which uses teacher-generated images rather than external real images as real samples for GAN training, providing a more attainable and informative adversarial target. Second, we adopt Step-Decoupled Parameterization, assigning independent model parameters to the two denoising steps to better match their distinct capacity demands. Third, we perform End-to-End Training with Iterative Regularization, allowing the first step to receive gradients from final image quality while preserving a meaningful intermediate generation through an explicit step-1 loss. Together, these designs substantially narrow the quality gap between 2-step and 8-step generation in both qualitative and quantitative evaluations, highlighting the potential of carefully tailored distillation strategies for improving the quality-efficiency trade-off in few-step generation.

Problem

• few-step generation
• 2-step image generation
• diffusion distillation
• high-fidelity generation
• model capacity

Innovation

• few-step diffusion distillation
• distribution-aligned adversarial learning
• step-decoupled parameterization
• end-to-end distillation
• image generation