FlashControl: One-Step Controllable Generator via Distillation-Friendly Single-Stream Teachers

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.22537 (cs) [Submitted on 15 Oct 2026] Title: FlashControl: One-Step Controllable Generator via Distillation-Friendly Single-Stream Teachers Authors: Ngan Nguyen, Dung Nguyen, Quan Dao, Dimitris N. Metaxas, Minh Hoai, Cuong Pham, Anh Tran.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.22537 (cs) [提交于 2026 年 10 月 15 日] 标题：FlashControl：通过易于蒸馏的单流教师模型实现一步可控生成。作者：Ngan Nguyen, Dung Nguyen, Quan Dao, Dimitris N. Metaxas, Minh Hoai, Cuong Pham, Anh Tran。

Abstract: Diffusion models have become a leading approach in visual generative modeling, particularly effective in text-to-image and controllable generation. Their main drawback is slow inference as it requires many sampling steps. While recent distillation methods enable one-step text-to-image generation, extending these techniques to controllable generation is still challenging.

摘要：扩散模型已成为视觉生成建模中的主流方法，在文本生成图像和可控生成方面尤为有效。其主要缺点是推理速度慢，因为需要大量的采样步骤。尽管近期的蒸馏方法实现了单步文本生成图像，但将这些技术扩展到可控生成领域仍然具有挑战性。

Existing controllable models, such as ControlNet, rely on dual-branch design that make distillation difficult. In this work, we adopt a simpler strategy. Instead of adding extra conditional branches, we finetune a selected subset of layers in the original text-to-image diffusion model to directly enable controllable generation.

现有的可控模型（如 ControlNet）依赖于双分支设计，这使得蒸馏变得困难。在这项工作中，我们采用了一种更简单的策略。我们没有添加额外的条件分支，而是对原始文本生成图像扩散模型中的选定层子集进行微调，以直接实现可控生成。

This results in DeltaControl, a single-flow, multi-step diffusion model capable of supporting spatial-condition generation while matching the performance of dual-branch approaches. Building on this teacher model, we further apply step-distillation to obtain FlashControl, a one-step model capable of spatial controllable generation with a single forward pass.

由此产生了 DeltaControl，这是一种单流、多步扩散模型，它能够在支持空间条件生成的同时，达到与双分支方法相当的性能。基于该教师模型，我们进一步应用步长蒸馏技术获得了 FlashControl，这是一种能够通过单次前向传递实现空间可控生成的单步模型。