SANA-Sprint: One-Step Diffusion with Continuous-Time Consistency Distillation

2025-03-12

AI Summary

This work addresses the high computational cost and slow inference of text-to-image (T2I) diffusion models. We propose SANA-Sprint, an efficient single-step diffusion model. Methodologically, it innovatively unifies training-free continuous-time consistency distillation (sCM) with latent adversarial distillation (LADD), enabling a unified, order-adaptive architecture natively integrated with ControlNet for real-time controllable generation. To our knowledge, this is the first method achieving single-step, high-fidelity 1024×1024 image synthesis: T2I latency is 0.1 s on H100 and 0.31 s on RTX 4090; ControlNet-guided generation takes 0.25 s on H100—10× faster than FLUX-schnell. Single-step generation achieves FID=7.59 and GenEval=0.74. Designed for consumer-grade real-time applications—including AI PCs—the framework significantly advances both generation efficiency and controllability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This paper presents SANA-Sprint, an efficient diffusion model for ultra-fast text-to-image (T2I) generation. SANA-Sprint is built on a pre-trained foundation model and augmented with hybrid distillation, dramatically reducing inference steps from 20 to 1-4. We introduce three key innovations: (1) We propose a training-free approach that transforms a pre-trained flow-matching model for continuous-time consistency distillation (sCM), eliminating costly training from scratch and achieving high training efficiency. Our hybrid distillation strategy combines sCM with latent adversarial distillation (LADD): sCM ensures alignment with the teacher model, while LADD enhances single-step generation fidelity. (2) SANA-Sprint is a unified step-adaptive model that achieves high-quality generation in 1-4 steps, eliminating step-specific training and improving efficiency. (3) We integrate ControlNet with SANA-Sprint for real-time interactive image generation, enabling instant visual feedback for user interaction. SANA-Sprint establishes a new Pareto frontier in speed-quality tradeoffs, achieving state-of-the-art performance with 7.59 FID and 0.74 GenEval in only 1 step - outperforming FLUX-schnell (7.94 FID / 0.71 GenEval) while being 10x faster (0.1s vs 1.1s on H100). It also achieves 0.1s (T2I) and 0.25s (ControlNet) latency for 1024 x 1024 images on H100, and 0.31s (T2I) on an RTX 4090, showcasing its exceptional efficiency and potential for AI-powered consumer applications (AIPC). Code and pre-trained models will be open-sourced.

Problem

• Efficient ultra-fast text-to-image generation
• Reducing inference steps from 20 to 1-4
• Real-time interactive image generation with ControlNet

Innovation

• Training-free continuous-time consistency distillation
• Unified step-adaptive model for 1-4 step generation
• Integration of ControlNet for real-time image generation