NumeriKontrol: Adding Numeric Control to Diffusion Transformers for Instruction-based Image Editing

TL;DR

NumeriKontrol introduces a numeric control framework for diffusion-based image editing, enabling precise, continuous attribute adjustments guided by natural language instructions with high fidelity and multi-condition support.

Contribution

It presents a novel Numeric Adapter for diffusion models, allowing precise numeric control in image editing, and introduces the CAT dataset for training and evaluation.

Findings

Achieves accurate and stable scale control across diverse editing tasks.

Supports zero-shot multi-condition editing with flexible instruction ordering.

Demonstrates high-quality, continuous attribute manipulation in experiments.

Abstract

Instruction-based image editing enables intuitive manipulation through natural language commands. However, text instructions alone often lack the precision required for fine-grained control over edit intensity. We introduce NumeriKontrol, a framework that allows users to precisely adjust image attributes using continuous scalar values with common units. NumeriKontrol encodes numeric editing scales via an effective Numeric Adapter and injects them into diffusion models in a plug-and-play manner. Thanks to a task-separated design, our approach supports zero-shot multi-condition editing, allowing users to specify multiple instructions in any order. To provide high-quality supervision, we synthesize precise training data from reliable sources, including high-fidelity rendering engines and DSLR cameras. Our Common Attribute Transform (CAT) dataset covers diverse attribute manipulations with accurate ground-truth scales, enabling NumeriKontrol to function as a simple yet powerful interactive editing studio. Extensive experiments show that NumeriKontrol delivers accurate, continuous, and stable scale control across a wide range of attribute editing scenarios. These contributions advance instruction-based image editing by enabling precise, scalable, and user-controllable image manipulation.