Training-Free Multi-Concept Image Editing

Niki Foteinopoulou; Ignas Budvytis; Stephan Liwicki

cs.CVFeb 24, 2026

Training-Free Multi-Concept Image Editing

Authors: Niki Foteinopoulou, Ignas Budvytis, Stephan Liwicki

Organizations: Cambridge Research Laboratory, Toshiba Europe, Cambridge, UK · Independent Researcher, UK

Abstract

Training-free image editing with diffusion models is highly desirable yet is complex and remains a significant challenge. While recent optimisation-based methods achieve strong zero-shot edits from text, they still struggle to preserve identity and capture intricate details, such as facial structure, surface texture, or object-specific geometry, that exist below the level of linguistic abstraction. To address this fundamental gap, we propose Concept Distillation Sampling (CDS). To the best of our knowledge, we are the first to introduce a unified, training-free framework for target-less, multi-concept image editing. CDS overcomes this linguistic bottleneck of previous methods by anchoring the editing process in the certainty of pretrained LoRA adapters. We integrate a highly stable distillation backbone (featuring ordered timesteps, regularisation, and negative-prompt guidance) with a novel dynamic weighting mechanism. This approach enables the composition and control of multiple visual concepts directly within the diffusion process, utilising spatially-aware priors from pretrained LoRA adapters without causing concept clashing. Our method preserves instance concept identity without requiring reference samples of the desired edit. Extensive quantitative and qualitative evaluations demonstrate that CDS establishes a new state-of-the-art over existing training-free editing and multi-LoRA composition methods on the InstructPix2Pix and ComposLoRA benchmarks. Project Page: https://nickyfot.github.io/cds/.

Figures & tables

Figure 1: Our method enables training-free, concept-based image editing with diffusion models. While recent state-of-the-art optimisation methods rely solely on text guidance and often lose instance identity during complex, multi-element changes, we unify a regularised optimisation objective with LoRA-driven concept composition. This enables multiple semantic or visual concepts ( e.g . facial structure, clothing, objects, or backgrounds captured in Dreambooth-style LoRAs) to be combined and controlled directly within the diffusion process. This approach preserves concept identity and fine-grained details that would be impossible to adequately describe in text alone, bridging the gap between text-based and visual concept-driven control.

Figure 2: Overview of our CDS image editing framework. We integrate multiple LoRA adapters into the image optimisation loop, enabling spatially-aware, concept-preserving edits across multiple concepts. The denoising trajectory follows ordered timesteps, allowing precise control over pose, style, and semantic attributes while maintaining overall composition.

Figure 3: Visualisation of timestep-ordered denoising. Comparison between the source image, $\nabla_{\theta}\mathcal{L}_{\mathrm{CDS}}$ at intermediate optimisation steps, and the final target image. Unlike DDS [ 8 ] , which samples timesteps uniformly at random, our method enforces a strict descending timestep order, enabling a coarse-to-fine denoising trajectory as is evident from the visualised gradients. Early steps capture high-frequency structural details such as edges, while later steps refine lower-frequency and stylistic components, resulting in a more coherent and stable edit.

Figure 4: Comparison of our optimisation objective vs baselines. Our method produces better results for the target subject with similarly perceived changes in the background.

	CLIPScore $\uparrow$	LPIPS $\downarrow$
DiffusionClip ${\ddagger}$	0.251 ± 0.022	0.572 ± 0.0594
PnP ${\ddagger}$	0.221 ± 0.036	0.310 ± 0.075
InstructPix2Pix ${\ddagger}$	0.219 ± 0.037	0.322 ± 0.215
DDS ${\ddagger}$	0.225 ± 0.031	0.104 ± 0.061
PDS†	0.298 ± 0.044	0.096 ± 0.033
Ours	0.308 ± 0.042 *	0.100 ± 0.042

Table 1: Quantitative Evaluation of our optimisation method against previous SoTA, on InstructPix2Pix. Our method achieves statistically significant improvements in CLIPScore while maintaining comparable LPIPS; the CLIPScore gain reflects a genuine semantic edge under a deliberate semantic–fidelity trade-off.

Table 2: Quantitative Evaluation of Concept Distillation Sampling (CDS). Performance comparison across four LoRA composition strategies for different numbers of LoRAs on ComposLoRA testbed. Bold values indicate statistical significance at the 99% confidence interval.

Panel of Figure 5 — Figure 5: Comparison of our CDS framework against previous LoRA-composition methods, evaluated using GPT-4V. Our method achieves higher perceived image and composition quality, with consistently higher pairwise win rates.

Figure 6: Examples of our method under different editing conditions. Our approach enables consistent and coherent edits across diverse scenarios, including pose changes, facial expression modifications, and element or object swaps ( e.g . , between different characters and environmental components). Each pair demonstrates the method’s ability to preserve subject and visual fidelity while adapting to the intended transformation.

	Avg. Rank $\downarrow$	Win Rate $\uparrow$
CDS	1.90	38%
Compose	2.62	31%
Switch	2.49	25%
Merge	3.00	5%

Table 3: Human evaluation comparing our CDS framework to previous LoRA-composition methods. In line with GPT-4V assessments, human raters consistently preferred our method, which achieved the lowest average rank and most frequent first-place selections (highest win rate).

	CLIPScore $\uparrow$	LPIPS $\downarrow$
DDS (200 steps)	0.225 ± 0.031	0.104 ± 0.061
DDS	0.305 ± 0.039	0.264 ± 0.061
Strict Timesteps (+DDS)	0.332 ± 0.032	0.461 ± 0.076
Regularisation (+DDS)	0.291 ± 0.041	0.081 ± 0.033
Negative Prompts (+DDS)	0.318 ± 0.040	0.319 ± 0.076
Full CDS	0.308 ± 0.042	0.100 ± 0.042

Table 4: CDS Core Objective Ablation. We establish a baseline with standard DDS (300 steps unless indicated) and iteratively introduce our objective components to isolate their effects. Our unified formulation achieves the optimal balance between executing semantic changes and maintaining structural integrity.

Figure 7: Samples generated from the base model using text descriptions. We observe distortions in the face, additional limbs, missing concepts, and inconsistent knowledge of characters and objects, all of which affect downstream tasks.

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 8: Sensitivity of the CDS objective to $\eta$ . When the intended semantic change is spatially small, the method is highly robust to large $\eta$ values.

$\eta$	CLIP $\uparrow$	LPIPS $\downarrow$
0.01	$0.316\pm 0.040$	$0.407\pm 0.151$
0.05	$0.308\pm 0.042$	$0.100\pm 0.042$
1	$0.304\pm 0.042$	$0.090\pm 0.047$
5	$0.299\pm 0.044$	$0.079\pm 0.034$
10	$0.299\pm 0.044$	$0.089\pm 0.035$

Appendix

Table 5: Sensitivity of the CDS objective to the

\eta

hyperparameter, evaluated on InstructPix2Pix.

	$N=2$	$N=3$	Avg.
$2\times 2$	$\mathbf{0.340\pm 0.030}$	$\mathbf{0.361\pm 0.022}$	$\mathbf{0.351}$
$4\times 4$	$0.335\pm 0.029$	$0.356\pm 0.022$	$0.346$
$8\times 8$	$0.329\pm 0.031$	$0.349\pm 0.020$	$0.339$
$16\times 16$	$0.322\pm 0.029$	$0.341\pm 0.022$	$0.332$

Appendix

Table 6: Effect of patch size on a subset of the ComposLoRA testbed. Although the method is relatively robust to patch size variations, a more localised approach (smaller patches) yields higher CLIPScores.

	$N=2$	$N=3$	Avg.
$\tau=0.002$	$\mathbf{0.340\pm 0.030}$	$\mathbf{0.361\pm 0.022}$	$\mathbf{0.351}$
$\tau=0.01$	$0.337\pm 0.030$	$0.360\pm 0.022$	$0.349$
$\tau=0.1$	$0.333\pm 0.029$	$0.360\pm 0.022$	$0.347$
$\tau=1$	$0.329\pm 0.029$	$0.359\pm 0.022$	$0.344$

Appendix

Table 7: Effect of

\tau

on a subset of the ComposLoRA testbed. Lower temperatures positively improve CLIPScore by sharpening concept isolation.

	CLIPScore $\uparrow$	LPIPS $\downarrow$
Merge	$0.273$	$0.38$
CDS (Ours)	$\mathbf{0.304}$	$\mathbf{0.21}$

Appendix

Table 8: SDXL results at

N=2

on a ComposLoRA-style protocol. CDS retains its identity-preservation and semantic-alignment advantages on a stronger backbone.

Panel of Figure 9 — Figure 9: Sample character + style change using SDXL. Merge collapses the target identity; CDS preserves source structure while introducing both concepts.

Figure 10: Examples of multi-concept edits on the anime subset of the ComposLoRA testbed for various $N$ .

Figure 11: Examples of multi-concept edits on the reality subset of the ComposLoRA testbed for various $N$ .

Panel of Figure 12 — Figure 12: Comparison of our CDS framework against previous LoRA-composition methods, evaluated using GPT-4V. Our method achieves consistently higher pairwise win rates across both subsets.

Table 9: Comparison of Concept Distillation Sampling (CDS) against previous LoRA-composition methods, evaluated using GPT-4V. Our method consistently achieves the highest total Image and Composition Quality scores.

Category	LoRA Name	Trigger Words	Source
Anime Style Subset
Character	Kamado Nezuko	kamado nezuko, black hair, pink eyes, forehead	Link
Character	Texas the Omertosa	omertosa, 1girl, wolf ears, long hair	Link
Character	Son Goku	son goku, spiked hair, muscular male, wristband	Link
Clothing	Garreg Mach Uniform	gmuniform, blue thighhighs, long sleeves	Link
Clothing	Zero Suit (Metroid)	zero suit, blue gloves, high heels	Link

Appendix

Table 10: Detailed descriptions of each LoRA in the ComposLoRA [ 41 ] Testbed.

Figure 13: Human Instructions for the Multi-Concept Editing task.

Figure 14: Example Human Evaluation Interface.

Training-Free Image Editing

See all 19 papers

GIDE: Unlocking Diffusion LLMs for Precise Training-Free Image Editing
Mar 22, 2026Zifeng Zhu, Jiaming Han, Jiaxiang Zhao +2Training-Free Image Editing Image Editing
Xi’an Jiaotong University · MMLab, The Chinese University of Hong Kong
Training-Free Image Editing with Visual Context Integration and Concept Alignment
Apr 6, 2026Rui Song, Guo-Hua Wang, Qing-Guo Chen +6Concept Alignment Training-Free Image Editing
HKUST · Alibaba International Digital Commerce Group · Institute for AI Industry Research (AIR), Tsinghua University +1
Unlocking the Potential of Image Editing via Concept Scaling and Dense Supervision
Aug 17, 2026Long Cui, Xiaoqian Liu, Qi Qin +4Image Editing Diffusion-Based Image Editing
Shanghai Jiao Tong University · Ant Group

Mar 22, 2026J/K move · Enter open · S save

GIDE: Unlocking Diffusion LLMs for Precise Training-Free Image Editing

Zifeng Zhu, Jiaming Han, Jiaxiang Zhao, Minnan Luo+1

Xi’an Jiaotong University · MMLab, The Chinese University of Hong Kong

Training-Free Image Editing Image Editing Diffusion-Based Image Editing Text-Guided Image Editing Diffusion Language Models Diffusion Model Inversion

Open paper AskPDF arXiv

While Diffusion Large Language Models (DLLMs) have demonstrated remarkable capabilities in multi-modal generation, performing precise, training-free image editing remains an open challenge. Unlike continuous diffusion models, the discrete tokenization inherent in DLLMs hinders the application of standard noise inversion techniques, often leading to structural degradation during editing. In this paper, we introduce GIDE (Grounded Inversion for DLLM Image Editing), a unified framework designed to bridge this gap. GIDE incorporates a novel Discrete Noise Inversion mechanism that accurately captures latent noise patterns within the discrete token space, ensuring high-fidelity reconstruction. We then decompose the editing pipeline into grounding, inversion, and refinement stages. This design enables GIDE supporting various editing instructions (text, point and box) and operations while strictly preserving the unedited background. Furthermore, to overcome the limitations of existing single-step evaluation protocols, we introduce GIDE-Bench, a rigorous benchmark comprising 805 compositional editing scenarios guided by diverse multi-modal inputs. Extensive experiments on GIDE-Bench demonstrate that GIDE significantly outperforms prior training-free methods, improving Semantic Correctness by 51.83% and Perceptual Quality by 50.39%. Additional evaluations on ImgEdit-Bench confirm its broad applicability, demonstrating consistent gains over trained baselines and yielding photorealistic consistency on par with leading models.

Abstract

Training-Free Image Editing

Explore similar work

GIDE: Unlocking Diffusion LLMs for Precise Training-Free Image Editing

Training-Free Image Editing with Visual Context Integration and Concept Alignment

Unlocking the Potential of Image Editing via Concept Scaling and Dense Supervision