FlowSlider: Training-Free Continuous Image Editing via Fidelity-Steering Decomposition

Taichi Endo; Guoqing Hao; Kazuhiko Sumi

cs.CVApr 2, 2026

FlowSlider: Training-Free Continuous Image Editing via Fidelity-Steering Decomposition

Authors: Taichi Endo, Guoqing Hao, Kazuhiko Sumi

Organizations: Aoyama Gakuin University

Abstract

Continuous image editing aims to provide slider-style control of edit strength while preserving source-image fidelity and maintaining a consistent edit direction. Existing learning-based slider methods typically rely on auxiliary modules trained with synthetic or proxy supervision. This introduces additional training overhead and couples slider behavior to the training distribution, which can reduce reliability under distribution shifts in edits or domains. We propose FlowSlider, a training-free method for continuous editing in Rectified Flow that requires no post-training. FlowSlider decomposes FlowEdit's update into (i) a fidelity term, which acts as a source-conditioned stabilizer that preserves identity and structure, and (ii) a steering term that drives semantic transition toward the target edit. Geometric analysis and empirical measurements show that these terms are approximately orthogonal, enabling stable strength control by scaling only the steering term while keeping the fidelity term unchanged. As a result, FlowSlider provides smooth and reliable control without post-training, improving continuous editing quality across diverse tasks.

Figures & tables

Figure 1: FlowSlider for continuous image editing : A slider-style editing framework that requires no learning and can be applied instantly to any edit. Simply specify a prompt pair to continuously control the editing intensity via the scaling parameter $s$ . Each row shows progressively stronger editing results starting from the real image.

Panel of Figure 2 — Figure 2: ‘Autumn $\rightarrow$ Winter’ editing example. Naive scaling in FlowEdit [ 23 ] introduces artifacts at $s=2$ , whereas FlowSlider maintains stable fidelity. Note that our FlowSlider is identical to FlowEdit when $s=1$ .

Panel of Figure 3 — Figure 3: Fidelity–steering decomposition and its implication for strength control. (a) Exact decomposition of the FlowEdit update into a same-state, cross-prompt steering term and a same-prompt, cross-state fidelity term: $V^{\Delta}=V_{\mathrm{fid}}+V_{\mathrm{steer}}$ . (b) Naive scaling ( $sV^{\Delta}$ ) magnifies both components and residual non-cancelled noise, which can drive the trajectory out of the source-consistent regime; decomposed scaling $V^{\Delta}_{s}=V_{\mathrm{fid}}+sV_{\mathrm{steer}}$ increases semantic strength while keeping source-conditioned stabilization unchanged.

Figure 4: Orthogonality between fidelity and steering components on our benchmark data. We plot the angle

\theta(t)=\arccos(\frac{\langle V_{\mathrm{fid}}(t),V_{\mathrm{steer}}(t)\rangle}{\|V_{\mathrm{fid}}(t)\|\,\|V_{\mathrm{steer}}(t)\|})

across timesteps and strength settings. Angles remain concentrated near

90^{\circ}

, indicating weak coupling between source-conditioned stabilization and semantic steering.

Panel of Figure 5 — Figure 5: Qualitative slider results. We compare our method against Kontinuous Kontext [ 30 ] and SliderEdit [ 49 ] . Each row shows the original (leftmost) and edited outputs at increasing slider strengths (left to right).

Method	Edit quality		Slider behavior
Method	CLIP-dir $\uparrow$	DreamSim $\downarrow$	Mono $\uparrow$	Smooth $\downarrow$
FlowEdit (CFG)	0.113	0.128	0.214	0.61
FlowEdit ( $n_{\max}$ )	0.151	0.266	0.530	0.61
FLUX-Kontext	0.470	0.269	0.324	4.02
Kontinuous Kontext	0.358	0.179	0.588	0.85
SliderEdit	0.457	0.337	0.391	0.27

Table 1: Quantitative evaluation results on our continuous-editing benchmark. We compare our method with other slider approaches along two dimensions: edit quality and slider behavior. An ideal slider method should achieve strong slider behavior while maintaining high edit quality.

Method	Edit quality		Slider behavior
Method	CLIP-dir $\uparrow$	DreamSim $\downarrow$	Mono $\uparrow$	Smooth $\downarrow$
FlowEdit (CFG)	0.063	0.045	0.350	1.64
FlowEdit ( $n_{\max}$ )	0.114	0.161	0.593	0.60
Kontinuous Kontext	0.245	0.188	0.545	0.27
SliderEdit	0.371	0.290	0.433	0.30
FlowSlider	0.286	0.093	0.672	0.11

Table 2: General editing evaluation on PIE-Bench [ 21 ] . This evaluation is included to demonstrate general performance beyond our continuous-editing benchmark. We report both edit quality and slider-behavior metrics, and FlowSlider shows trends consistent with those on our continuous-editing benchmark.

Method	Edit quality		Slider behavior
Method	CLIP-dir $\uparrow$	DreamSim $\downarrow$	Mono $\uparrow$	Smooth $\downarrow$
Naive scaling	0.255	0.423	0.371	0.31
Linear Interpolation	0.119	0.342	0.417	0.04
FlowSlider (Ours)	0.400	0.090	0.833	0.01

Table 3: Quantitative ablation results. We compare two training-free alternatives—naive scaling of the FlowEdit update (

s\,V^{\Delta}

) and linear velocity interpolation—against FlowSlider. The evaluation is conducted under the same setting as Tab. 1 .

Panel of Figure 6 — Figure 6: Ablation on strength modulation. Naive scaling of the FlowEdit update and linear velocity interpolation lead to visible artifacts and degraded structure for $s>1$ . FlowSlider remains stable across strengths by keeping the fidelity term fixed and scaling only the steering term.

Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

Category	Editing tasks	Samples
Scene/weather/time	season (49), weather (21), time of day (35)	105
Style	style change (44)	44
Degradation	time decay (38)	38
Color	color change (18)	18
Portrait Attributes	expression, hair color, makeup, aging, etc.	45
Total		250

Appendix

Table 4: Benchmark specification. The dataset contains 250 samples spanning 5 categories and 13 editing tasks.

Panel of Figure 7 — Figure 7: Representative benchmark samples. Two examples from each of the five categories with abbreviated source–target prompt pairs. The benchmark contains 250 samples spanning 13 editing tasks (see Tab. 4 ).

Panel of Figure 8 — Figure 8: Additional qualitative comparisons (1/4). Each row shows the original (leftmost) and edited outputs at increasing slider strengths (left to right). Images are center-cropped to square for display.

Panel of Figure 9 — Figure 9: Additional qualitative comparisons (2/4). Same layout as Fig. 8 .

Panel of Figure 10 — Figure 10: Additional qualitative comparisons (3/4). Same layout as Fig. 8 .

Panel of Figure 11 — Figure 11: Additional qualitative comparisons (4/4). Same layout as Fig. 8 .

Figure 12: Fidelity–edit strength trade-off. CLIP-T (edit effect) vs. four preservation metrics across slider strengths. FlowSlider achieves higher edit effect while better preserving source consistency compared to Kontinuous Kontext [ 30 ] and SliderEdit [ 49 ] , occupying the upper-left (better) region in all four panels.

Panel of Figure 13 — Figure 13: Failure at extreme strengths. A gradual-editing task (decay) at increasing $s$ . Results remain plausible up to $s=4$ , but structural collapse and over-saturation emerge at larger values.

Panel of Figure 14 — Figure 14: Early saturation for discrete concept change. The concept “rabbit $\to$ cat” is fully realized at $s{=}1$ . Further scaling introduces over-saturation and color artifacts rather than meaningful editing.

Panel of Figure 15 — Figure 15: Bidirectional editing with negative strength. Setting $s<0$ reverses the edit direction, enabling bidirectional control from a single prompt pair without retraining. Each row shows the same source image edited with negative (left) and positive (right) strengths.

Training-Free Image Editing

See all 19 papers

SliderEdit++: Improved Slider Trajectories in Instruction-Based Image Editing
Oct 9, 2026Arman Zarei, Sayan Nag, Bryan Wang +5Image Editing Text-Guided Image Editing
Adobe Research · University of Maryland
UniSlider: Perceptually Uniform Sliders for Continuous Image Editing
Oct 5, 2026David Serrano-Lozano, Duygu Ceylan, Yannick Hold-Geoffroy +3Image Editing Evaluation Diffusion-Based Image Editing
Computer Vision Center · Universitat Autònoma de Barcelona · Adobe Research
VeloEdit: Training-Free Consistent and Continuous Instruction-Based Image Editing via Velocity Field Decomposition
Mar 11, 2026Zongqing Li, Zhihui Liu, Yujie Xie +3Training-Free Image Editing Image Editing
Xiamen University · Truesight

Oct 9, 2026J/K move · Enter open · S save

SliderEdit++: Improved Slider Trajectories in Instruction-Based Image Editing

Arman Zarei, Sayan Nag, Bryan Wang, Franck Dernoncourt+4

Adobe Research · University of Maryland

Image Editing Text-Guided Image Editing Image Editing Evaluation

Open paper AskPDF arXiv

Slider-based control aims to give users continuous, fine-grained control over the intensity of individual edits in instruction-based image editing. However, existing approaches fail to deliver reliable continuous controllability: traversing a slider can produce abrupt visual jumps and non-monotonic changes, while jointly controlled sliders can interfere with one another. To systematically study these limitations, we introduce SlideBench, a comprehensive benchmark for continuous image editing spanning diverse editing categories and equipped with a unified suite of metrics quantifying monotonic progression, perceptual continuity, and inter-slider disentanglement. Building on these observations, we propose SliderEdit++, which explicitly regularizes the intermediate editing trajectory rather than supervising only its endpoints. Our method combines velocity- and semantic-space trajectory regularization with spectral control to encourage smooth, predictable slider traversal, together with a disentanglement objective that reduces interference between independently controlled edits. On SlideBench, SliderEdit++ achieves state-of-the-art performance, improving the overall score by 18.5% relative to the strongest baseline. At the same time, using a few-step editing backbone makes our method approximately 3x to 20x faster than existing slider-based approaches, enabling substantially more responsive interactive editing.

Abstract

Training-Free Image Editing

Explore similar work

SliderEdit++: Improved Slider Trajectories in Instruction-Based Image Editing

UniSlider: Perceptually Uniform Sliders for Continuous Image Editing

VeloEdit: Training-Free Consistent and Continuous Instruction-Based Image Editing via Velocity Field Decomposition