One Evaluation Is Enough: Inversion-Free Image Editing with Trajectory Correction

Lifan Jiang; Taiyang Chen; Senlin cheng; Bolei Xu; Xiaobo Li; Can Ye; Dehai Zhao; Deng Cai

cs.CVOct 22, 2026

One Evaluation Is Enough: Inversion-Free Image Editing with Trajectory Correction

Authors: Lifan Jiang, Taiyang Chen, Senlin cheng, Bolei Xu, Xiaobo Li, Can Ye, Dehai Zhao, Deng Cai,

Organizations: State Key Lab of CAD&CG, Zhejiang University · Ant Group · Chongqing Key Laboratory of Image Cognition, Chongqing University of Posts and Telecommunications

Abstract

Real-image editing commonly relies on inversion and iterative sampling, while efficient one-step editors often stabilize a large update by estimating and averaging editing fields at multiple noise levels. We ask a stricter question: is a single temporal field evaluation sufficient for reliable inversion-free editing? Our answer rests on a previously underused duality in a variance-preserving (VP) one-step model: one source-target query at a shared noisy state simultaneously determines a local angular semantic tangent and a clean endpoint displacement. We integrate the tangent over the analytically known remaining VP angle to obtain a nominal one-step transport, and use the endpoint displacement to correct its long-horizon extrapolation through a constrained projection. This transport-first, correction-second construction yields Endpoint-Calibrated Transport (ECT), a closed-form, training-free editor requiring only one noise level, two text conditions packed into one denoiser batch, and one latent update. ECT requires no image inversion, trajectory integration, second temporal observation, or iterative optimization, and can be directly applied to pretrained epsilon-prediction VP one-step generators. Across the reported benchmark settings, ECT delivers stable gains in editing quality within an extremely short editing time.

Figures & tables

Figure 1: Edits with ECT (ours), ChordEdit, and FlowEdit, covering object addition and removal, attribute and background changes, stylization, and object replacement.

Figure 2: Editing pipelines, illustrated by a horse-to-zebra edit. (a) FlowEdit accumulates source–target field differences along a multi-step path. (b) ChordEdit combines residuals at $t$ and $t-\delta$ into one update, using two temporal supports and four condition–time branches. (c) ECT uses only $R_{t}(z_{t})$ from two text conditions at one shared noisy state. The same predictions determine the angular residual $g_{t}$ and endpoint displacement $d_{t}$ : angular transport gives $\widehat{a}_{t}=-\theta_{t}g_{t}$ , and endpoint calibration gives $u_{t}^{*}=\widehat{a}_{t}-\zeta d_{t}$ . A single update $x_{\mathrm{edit}}=x_{\mathrm{src}}+\lambda u_{t}^{*}$ then produces the edit (Equation 22 ). The sketches are schematic; the update rescales a shared edit axis.

Figure 3: Qualitative comparisons of one-step, few-step, and multi-step editors. Rows show flower addition, crystal-ball replacement, pixel-art stylization, and robot transformation. Source images and instructions illustrate the preservation–alignment trade-offs.

Figure 4: Schematic motivation for single-step calibration. (a) Simple drift produces dispersed endpoints. (b) ChordEdit combines two residuals. (c) Coherent updates can still miss the target distribution. (d) Transport and correction calibrate the endpoints. Direction changes abstract high-dimensional behavior; the exact ECT update is collinear with $d_{t}$ (Equation 21 ).

Figure 5: Qualitative effects of angular transport on two representative cases. From left to right: source, result without transport and its enlarged crop, then the “w/ Transport” result and its enlarged crop.

Figure 6: User-study first-place shares (%), based on mean participant scores per case, with no top ties. ECT leads in text alignment ( $38.6\%$ ), structure preservation ( $33.8\%$ ), and aesthetics ( $34.0\%$ ).

Figure 7: Preservation–alignment trade-offs on PIE-Bench. Lower LPIPS and higher CLIP are preferable; LPIPS is unscaled. (a) $t=0.75$ gives lower LPIPS at comparable CLIP than the other plotted evaluation times. (b) $\zeta=0.7$ gives the most favorable plotted trade-off.

Model	Setting	Time (s) $\downarrow$	CPU (%) $\downarrow$	Mem. (MiB) $\downarrow$	CP $\uparrow$	VQ $\uparrow$	PC $\uparrow$
SB-v2	ChordEdit	0.265	357	1298.74	3.91	4.37	4.05
SB-v2	ECT (2-step)	0.259	446	1382.99	3.99	4.42	4.10
SB-v2	ECT (1-step)	0.213 -0.5pt	190 -0.5pt	1075.93 -0.5pt	4.03	4.45	4.13
SD-T	ChordEdit	0.265	357	1297.07	3.95	4.37	4.00
SD-T	ECT (2-step)	0.259	452	1386.67	4.04	4.44	4.07
SD-T	ECT (1-step)	0.212 -0.5pt	190 -0.5pt	1075.96 -0.5pt	4.09	4.65	4.31

Table 3: Ablations on PIE-Bench: step count (a), angular transport (b), and extra NFE (c). Blue marks ECT variants, bold better values, and orange resource reductions.

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

Metric	Direction	Input	Primary interpretation
MSE	$\downarrow$	Masked source/output	Mean pixel error
PSNR	$\uparrow$	Masked source/output	Log pixel fidelity
SSIM	$\uparrow$	Masked source/output	Local structure preservation
LPIPS	$\downarrow$	Masked source/output	Learned perceptual distance
CLIP	$\uparrow$	Full output/instruction	Target semantic alignment

Appendix

Table 4: Automatic image metrics and their interpretation.

Column	Statistic	Scope	Unit
Time	Arithmetic mean of $d_{i}$	Pipeline call	s/image
CPU	Time-weighted process utilization	Inference windows	core-%
Mem.	Time-weighted process RSS	Inference windows	MiB
VRAM	Peak PyTorch reserved memory	Measured run	MiB

Appendix

Table 5: Resource metrics reported in the main tables.

Score 1: poor	Score 3: adequate	Score 5: excellent
CP: Content and Identity Preservation
Severe unintended loss, deformation, or identity/layout change.	Source remains recognizable, with visible unintended local changes.	Unrequested identity, geometry, layout, and detail are preserved.
VQ: Visual Quality and Artifact Control
Severe blur, noise, corruption, or deformation.	Usable output with visible local blur, noise, or edge artifacts.	Sharp, detailed output without visible editing artifacts.
PC: Physical and Perceptual Coherence
Severe conflicts in lighting, perspective, anatomy, scale, depth, material, or style.	Broadly plausible, with minor physical or perceptual inconsistency.	Edited content is naturally integrated and internally coherent.

Appendix

Table 6: ImgEdit-Judge scoring anchors.

Dimension	Rating criterion
Text alignment	Agreement between the edited image and instruction
Structure preservation	Preservation of non-target content, identity, geometry, and layout
Overall aesthetics	Naturalness, detail, consistency, and artifact control

Appendix

Table 7: Human rating dimensions.

Figure 8: Additional transport ablations complementing Figure 5 . From left to right: source, result without transport and its crop, then the “w/ Transport” result and its crop. The highlighted regions show an extra shaft and altered hand–shoulder geometry without transport, reduced with the “w/ Transport” result.

Figure 9: Six qualitative comparisons on PIE-Bench, with source, ECT (ours), ChordEdit, and FlowEdit in each case. Left: object replacement and two style edits. Right: a robot transformation and two object additions. The examples show differences in edit extent and source preservation, including incomplete spatial placement in the flowers-in-mouth case.

	Consistency (background preservation)
Seed	PSNR $\uparrow$	MSE $\downarrow$	LPIPS $\downarrow$	SSIM $\uparrow$	CLIP $\uparrow$
1	23.428	6.641	112.832	77.825	30.23
20	23.598	6.435	110.950	77.834	30.43
64	23.550	6.447	111.065	77.816	30.28
100	23.617	6.452	111.691	77.775	30.34
512	23.548	6.513	112.784	77.888	30.42

Appendix

Table 8: Seed stability on PIE-Bench (SD-Turbo). Eleven independent seeds, all other settings fixed. The standard deviations across seeds are about an order of magnitude smaller than the ECT–ChordEdit gap on every preservation metric, and the worst seed still outperforms matched ChordEdit on all four. The default seed is shaded blue; the reference row is the matched ChordEdit entry of Table 2 .

	VP schedule				ImgEdit-Judge
Setting	$\alpha_{\mathrm{prox}}$	$\sigma_{\mathrm{prox}}$	$\theta_{\mathrm{prox}}/\sigma_{\mathrm{prox}}$	$\theta_{\mathrm{prox}}/\sigma_{\mathrm{prox}}-\zeta$	CP $\uparrow$	VQ $\uparrow$	PC $\uparrow$
$t_{\mathrm{prox}}=0.1$	0.945634	0.325234	1.019	0.319	2.92	4.33	3.99
$t_{\mathrm{prox}}=0.3$	0.768440	0.639921	1.085	0.385	4.04	4.50	4.18
$t_{\mathrm{prox}}=0.5$	0.525673	0.850686	1.196	0.496	3.48	4.35	3.96
$t_{\mathrm{prox}}=0.7$	0.288320	0.957534	1.335	0.635	3.30	4.15	3.59
Ref. : ChordEdit (SD-T)	—	—	—	—	3.95	4.37	4.00

Appendix

Table 9: Supplementary extra-NFE refinement on PIE-Bench (SD-Turbo). This variant is not used in the main results. Among the tested refinement levels,

t_{\mathrm{prox}}=0.3

(blue) has the highest CP, VQ, and PC. ChordEdit is a reference, not an equal-NFE control. Best value per column in bold.

Setting	PSNR $\uparrow$	MSE $\downarrow$	LPIPS $\downarrow$	SSIM $\uparrow$	CLIP $\uparrow$
	Consistency (background preservation)
$\lambda=1$ (ours)	23.740	6.220	109.040	78.100	30.20
$\lambda=2$	21.050	11.655	146.676	74.432	30.70
$\lambda=3$	19.084	18.150	183.486	71.209	30.60
$\lambda=4$	17.724	24.364	213.623	68.508	30.20
$\lambda=5$	16.717	30.136	239.572	66.179	29.90

Appendix

Table 10: Effect of the edit-strength coefficient

\lambda

on PIE-Bench (SD-Turbo). Preservation degrades monotonically with

\lambda

; CLIP peaks at

\lambda=2

and then declines. Its

0.50

-point gain costs

2.690

dB PSNR and

37.636

scaled LPIPS units. Blue marks

\lambda=1

, the preservation-oriented default.

Diffusion-Based Image Editing

Diffusion-based image editing changes image content, style, or attributes while aiming to preserve the parts users want to keep.

See all 225 papers

PreciseEdit: Strong Yet Stable Semantic Transport for One-Step Image Editing
Oct 16, 2026Jingyi Wang, Xiao LuImage Editing Text-Guided Image Editing
Hainan University, School of Computer Science and Technology, Haikou, China · Shanghai University, School of Mechatronic Engineering and Automation, Shanghai, China
ChordEdit: One-Step Low-Energy Transport for Image Editing
Feb 22, 2026Liangsi Lu, Xuhang Chen, Minzhe Guo +3Text-Guided Image Editing Diffusion-Based Image Editing
Guangdong University of Technology · Huizhou University · Shenzhen University +1
Training-free image inversion for one-step diffusion models
May 31, 2026Tao Wu, Senmao Li, Yaxing Wang +3Diffusion Model Inversion Text-Guided Image Editing
Computer Vision Center, Universitat Autonoma de Barcelona, Barcelona, 08193, Spain · Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), Abu Dhabi, United Arab Emirates · Jilin University, 130012, Changchun, Jilin, China +3

Oct 16, 2026J/K move · Enter open · S save

PreciseEdit: Strong Yet Stable Semantic Transport for One-Step Image Editing

Jingyi Wang, Xiao Lu

Hainan University, School of Computer Science and Technology, Haikou, China · Shanghai University, School of Mechatronic Engineering and Automation, Shanghai, China

Image Editing Text-Guided Image Editing Diffusion-Based Image Editing Training-Free Image Editing

Open paper AskPDF arXiv

Recent one-step text-to-image (T2I) models enable efficient image synthesis, opening up new possibilities for interactive Web-based image editing. However, existing one-step editing methods often construct a global transport field through simple vector operations between conditional predictions. While such strong transport can effectively drive the image toward the target semantics, it may be released along unreliable directions or over non-edited regions, leading to structural distortions within the editing region and background drift elsewhere. More importantly, stabilizing the editing process through global smoothing or uniformly weakening the transport field often makes large semantic changes overly conservative. In this work, we revisit one-step image editing from the perspective of the tension between semantic strength and transport stability, and identify a key insight: strong semantic transport is not inherently unstable. What matters is whether it is applied along reliable directions and at reliable spatial locations. Motivated by this insight, we introduce PreciseEdit, a model-agnostic, training-free and inversion-free one-step editing framework that strengthens target-semantic alignment while preserving background fidelity and structural consistency. We further propose a dual-noise-scale role decomposition strategy: a higher-noise scale produces the strong target-directed transport used for actual editing, whereas a lower-noise scale jointly serves as a probe for directional agreement and spatial structure. Extensive experiments on PIE-Bench demonstrate that PreciseEdit achieves state-of-the-art performance and effectiveness on various editing tasks.

Abstract

Diffusion-Based Image Editing

Explore similar work

PreciseEdit: Strong Yet Stable Semantic Transport for One-Step Image Editing

ChordEdit: One-Step Low-Energy Transport for Image Editing

Training-free image inversion for one-step diffusion models