ChordEdit: One-Step Low-Energy Transport for Image Editing
arXiv - 2026
The ChordEdit paper describes one-step low-energy transport for image editing and exposes the demo parameters mirrored by this tool.
Loading...
Input
Upload a source image and describe its content and the desired result in English.
Image formats
Output
One PNG edit generated from the uploaded image through ChordEdit plus SD-Turbo.
Use case
One-step semantic image edits while preserving the original structure as much as possible
This tool is under maintenance. Please try again later.
Include the tool name, steps and the problem in your feedback. Screenshots can help us locate it.
ChordEdit edits an uploaded image in one low-energy transport step, using source and target prompts with SD-Turbo components.
ChordEdit is a training-free image editing method built around SD-Turbo component weights and an official one-step transport pipeline.
arXiv - 2026
The ChordEdit paper describes one-step low-energy transport for image editing and exposes the demo parameters mirrored by this tool.
Includes public test results for models used here and elsewhere. The complete source table is shown by default; conditions and metrics can also be selected for comparison. Scores from different test conditions are not directly comparable.
Only records with verified public inference implementations and matching weights are shown; algorithms requiring no weights are also eligible. Closed services, unreleased weights and unmatched experiment versions are excluded. This is not an all-model leaderboard.
Public weights do not establish a fully open-source license. Expand the data for declared weight licenses and access conditions. Undeclared licenses are marked separately.
Choose a dataset to view all metrics for each model. Rows follow the official rank or selected metric. Bold marks each column’s best comparable value, including ties. Different evaluation protocols are ranked separately; missing scores are shown as -.
Deployable entries 10 / source rows 12
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
InfEdit (SD1.4) · Few-step (4 steps) · ChordEdit (SwiftBrush-v2) · One-step
Paper Table 1 rows on PIE-bench. Higher PSNR and CLIP-Edited are better; runtime and VRAM are source-reported context, not a TelkNet re-run.
| Metric rank | Model | Test setting | Consistency | CLIP Semantics | Properties | Efficiency | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PSNR ↑ | MSE 10 3 ↓ | LPIPS 10 3 ↓ | Whole ↑ | Edited ↑ | T-free | I-free | Step ↓ | NFE ↓ | Runtime(s) ↓ | VRAM(MiB) ↓ | |||
| 1 | ChordEdit (w/o prox, SD-Turbo) | One-step | 23.89 | 5.05 | 88.36 | 24.97 | 21.87 | ✓ | ✓ | 1 | 1 | 0.20 | 6988 |
| 2 | InstantEdit (PeRFlow-SD1.5)Public weights · Inference codePeRFlow DreamShaper SD1.5Canny ControlNet SD1.5Inference code hansyan/perflow-sd15-dreamshaper@87a2d778a21cd6c9079217a53aea464d5c64c1ccWeight license (repository declaration): cc-by-nc-4.0 | Few-step (4 steps) | 23.80 | 4.21 | 60.92 | 24.97 | 21.82 | ✓ | ✗ | 4 | 8 | 1.30 | 16270 |
| 3 | ChordEdit (SD-Turbo)Used on this site | One-step | 22.20 | 6.84 | 128.25 | 25.58 | 22.96 | ✓ | ✓ | 1 | 2 | 0.38 | 6988 |
| 4 | FlowEdit (SD3)Public weights · Inference codestabilityai/stable-diffusion-3-medium-diffusers@ea42f8cef0f178587cf766dc8129abd379c90671Inference codeWeight license not declaredTerms acceptance or access request required | Multi-step ( ≥ 30 steps) | 22.17 | 7.69 | 104.81 | 26.64 | 23.69 | ✓ | ✓ | 33 | 33 | 7.22 | 17140 |
| 5 | Direct Inversion + MasaCtrl | Multi-step ( ≥ 30 steps) | 21.78 | 7.99 | 87.38 | 24.42 | 21.38 | ✓ | ✗ | 50 | 100 | 79.10 | 12272 |
| 6 | SwiftEdit (SwiftBrush-v2)Public weights · Inference code | One-step | 21.71 | 8.22 | 91.22 | 24.93 | 21.85 | ✗ | ✗ | 1 | 2 | 0.54 | 15060 |
| 7 | TurboEdit (SDXL-Turbo) | Few-step (4 steps) | 21.44 | 9.49 | 108.60 | 24.66 | 21.79 | ✓ | ✓ | 4 | 4 | 2.69 | 13826 |
| 8 | Direct Inversion + PnP | Multi-step ( ≥ 30 steps) | 21.43 | 8.10 | 106.26 | 25.48 | 22.63 | ✓ | ✗ | 50 | 100 | 28.03 | 9262 |
| 9 | DDIM + PnP | Multi-step ( ≥ 30 steps) | 21.26 | 8.42 | 113.58 | 25.45 | 22.54 | ✓ | ✗ | 50 | 100 | 28.01 | 9262 |
| 10 | DDIM + MasaCtrl | Multi-step ( ≥ 30 steps) | 21.25 | 8.58 | 106.59 | 24.13 | 21.13 | ✓ | ✗ | 50 | 100 | 55.20 | 12272 |
Markers identify models or components used here. Scores and settings come from the cited tests, not a new evaluation of the complete site pipeline.
Datasets, versions and conditions remain separate, including sources with no eligible ranking rows.
Deployable entries 10 / source rows 12
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
InfEdit (SD1.4) · Few-step (4 steps) · ChordEdit (SwiftBrush-v2) · One-step
Scores apply to the listed dataset, metric and model version; results from different tests are not directly comparable.
Choose a dataset to view all metrics for each model. Rows follow the official rank or selected metric. Bold marks each column’s best comparable value, including ties. Different evaluation protocols are ranked separately; missing scores are shown as -.
Deployable entries 10 / source rows 12
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
InfEdit (SD1.4) · Few-step (4 steps) · ChordEdit (SwiftBrush-v2) · One-step
Paper Table 1 rows on PIE-bench. Higher PSNR and CLIP-Edited are better; runtime and VRAM are source-reported context, not a TelkNet re-run.
| Metric rank | Model | Test setting | Consistency | CLIP Semantics | Properties | Efficiency | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PSNR ↑ | MSE 10 3 ↓ | LPIPS 10 3 ↓ | Whole ↑ | Edited ↑ | T-free | I-free | Step ↓ | NFE ↓ | Runtime(s) ↓ | VRAM(MiB) ↓ | |||
| 1 | ChordEdit (w/o prox, SD-Turbo) | One-step | 23.89 | 5.05 | 88.36 | 24.97 | 21.87 | ✓ | ✓ | 1 | 1 | 0.20 | 6988 |
| 2 | InstantEdit (PeRFlow-SD1.5)Public weights · Inference codePeRFlow DreamShaper SD1.5Canny ControlNet SD1.5Inference code hansyan/perflow-sd15-dreamshaper@87a2d778a21cd6c9079217a53aea464d5c64c1ccWeight license (repository declaration): cc-by-nc-4.0 | Few-step (4 steps) | 23.80 | 4.21 | 60.92 | 24.97 | 21.82 | ✓ | ✗ | 4 | 8 | 1.30 | 16270 |
| 3 | ChordEdit (SD-Turbo) | One-step | 22.20 | 6.84 | 128.25 | 25.58 | 22.96 | ✓ | ✓ | 1 | 2 | 0.38 | 6988 |
| 4 | FlowEdit (SD3)Public weights · Inference codestabilityai/stable-diffusion-3-medium-diffusers@ea42f8cef0f178587cf766dc8129abd379c90671Inference codeWeight license not declaredTerms acceptance or access request required | Multi-step ( ≥ 30 steps) | 22.17 | 7.69 | 104.81 | 26.64 | 23.69 | ✓ | ✓ | 33 | 33 | 7.22 | 17140 |
| 5 | Direct Inversion + MasaCtrl | Multi-step ( ≥ 30 steps) | 21.78 | 7.99 | 87.38 | 24.42 | 21.38 | ✓ | ✗ | 50 | 100 | 79.10 | 12272 |
| 6 | SwiftEdit (SwiftBrush-v2)Public weights · Inference code | One-step | 21.71 | 8.22 | 91.22 | 24.93 | 21.85 | ✗ | ✗ | 1 | 2 | 0.54 | 15060 |
| 7 | TurboEdit (SDXL-Turbo) | Few-step (4 steps) | 21.44 | 9.49 | 108.60 | 24.66 | 21.79 | ✓ | ✓ | 4 | 4 | 2.69 | 13826 |
| 8 | Direct Inversion + PnP | Multi-step ( ≥ 30 steps) | 21.43 | 8.10 | 106.26 | 25.48 | 22.63 | ✓ | ✗ | 50 | 100 | 28.03 | 9262 |
| 9 | DDIM + PnP | Multi-step ( ≥ 30 steps) | 21.26 | 8.42 | 113.58 | 25.45 | 22.54 | ✓ | ✗ | 50 | 100 | 28.01 | 9262 |
| 10 | DDIM + MasaCtrl | Multi-step ( ≥ 30 steps) | 21.25 | 8.58 | 106.59 | 24.13 | 21.13 | ✓ | ✗ | 50 | 100 | 55.20 | 12272 |
Find processing methods and steps suited to your material.