Research map / Image generation / Editing and style transfer
DirectEdit: Step-Level Accurate Inversion for Flow-Based Image Editing
Desong Yang, Mang Ye
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Direct Alignment instead of stepwise correction (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | Structure Distance: 29.22 -> 17.94 | Improved |
| add Direct Alignment instead of stepwise correction (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | background PSNR: 31.12 -> 32.63 | Improved |
| add Direct Alignment (residual alignment of forward path) | Euler inversion on FLUX.1-dev · PIE-Bench reconstruction | step-level MSE (avg): 0.2857 -> 0.0006 | Improved |
| add Direct Alignment | inversion NFE cost · PIE-Bench reconstruction | NFE: 60 -> 60 | No clear effect |
| add attention feature injection (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | background PSNR: 31.93 -> 32.63 | Improved |
| add attention feature injection (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | CLIP similarity (edited): 22.69 -> 22.45 | Hurt |
| add mask-guided noise blending (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | background PSNR: 24.70 -> 32.63 | Improved |
| add mask-guided noise blending (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | CLIP similarity (edited): 22.74 -> 22.45 | Hurt |
| use multi-branch mask instead of rectangular bounding box (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | background PSNR: 27.86 -> 32.63 | Improved |
| add Direct Alignment over vanilla Euler inversion (ablation) | DirectEdit (FLUX.1-dev) · PIE-Bench | background PSNR: 16.81 -> 32.63 | Improved |
| increase attention injection steps t_inj (ablation) | DirectEdit · PIE-Bench qualitative | edit vs source similarity trade-off | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- DNAEdit — DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing (NeurIPS 2025) · better on 5 of 6 shared results
- FireFlow — FireFlow: Fast Inversion of Rectified Flow for Image Semantic Editing (ICML 2025) · better on 1 of 2 shared results
- FlowEdit — FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models (ICCV 2025) · better on 2 of 2 shared results
- FlowAlign — FlowAlign: Trajectory-Regularized, Inversion-Free Flow-based Image Editing (ICLR 2026) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Inversion-Free Image Editing with Language-Guided Diffusion Models (CVPR 2024 · similarity 0.90)
- Unveil Inversion and Invariance in Flow Transformer for Versatile Image Editing (CVPR 2025 · similarity 0.90)
- PostEdit: Posterior Sampling for Efficient Zero-Shot Image Editing (ICLR 2025 · similarity 0.90)
- Tuning-Free Inversion-Enhanced Control for Consistent Image Editing (AAAI 2024 · similarity 0.90)
- FireFlow: Fast Inversion of Rectified Flow for Image Semantic Editing (ICML 2025 · similarity 0.89)
- SplitFlow: Flow Decomposition for Inversion-Free Text-to-Image Editing (NeurIPS 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
