Inversion-free method evaluated for high-dimensional image counterfactuals
On CelebA and ADNI, the authors compare inversion-based and inversion-free methods on effectiveness and identity preservation.
Stories per day · 30 days
2 RSS
Relatedbenchmark2benchmarks2diffusion2distillation2multimodal2multi-turn2on-device2self-forcing2acedit1adni1adobe1agents1ai-agents1asymmetric-conditioning1autoregressive1causal-graphs1celeba1clip1
36 stories, newest first · one card per story
Stories tagged #image-editing
TailSteer Improves Multi-turn Image Editing Robustness
TailSteer reduces cumulative drift in multi-turn editing while preserving base model performance.
Adobe develops EditAgent for professional image editing
EditAgent generates parallel variants and uses verification data to outperform fine-tuned models on 600-edit benchmarks.
EditAgent expands tools and verification for agentic image editing
Experts preferred parallel listwise judging over sequential refinement in 78.5% of decided comparisons.
MoI-Bench scores mobile image generation and editing gaps
The suite has 64 subtasks and 1,150 samples, with MoI-Eval scoring instruction following, quality and detail preservation.
PREEN predicts editor presets for exemplar color transfer
The model regresses 29 smartphone-editor sliders, runs at sub-second on-device latency, and leads PST50 fidelity metrics.
MT Forcing trains multi-turn image editing to limit drift
Single-Turn DMD and Teacher-Forcing Guidance target artifact buildup and unintended changes across edit turns.
LayerEditor generates reusable RGBA layers during instruction-guided image edits
The method distills a pretrained editing prior into both the canvas and layers composed from LayerEditor-100K.
MTEdit dataset trains multi-turn image editing without rollout drift
A 500K-trajectory dataset and MTEdit-Bench back a fine-tune that leads evaluated open-source editors.
ACEdit: Stable Multi-Turn Image Editing via Asymmetric Conditioning
ACEdit stabilizes multi-turn image editing by distilling a teacher while the student edits its own outputs.
Krea 3 – Frontier Image Model 4× faster than Nano Banana Pro
Krea launched Krea 3, an image editor it says is faster and cheaper than Nano Banana Pro.
ACEdit uses asymmetric conditioning to stabilize multi-turn image edits
The student edits its own prior outputs while a teacher stays on the clean image with a compound instruction.
SliderEdit++ regularizes slider paths for smoother image edits
On SlideBench the paper reports an 18.5% higher overall score and about 3x to 20x faster editing.
InfoVis: Unified Slide Image Generation and Editing
InfoVis, a 9B slide model, nearly matches Nano-Banana-2 on a 1,527-instance benchmark.
Qwen/Qwen-Image-2.1-Turbo — new model trending #30 on Hugging Face
Qwen released Qwen-Image-2.1-Turbo, an eight-step accelerated 7B text-to-image checkpoint.
5
ThinkDrag: Semantic Drag-Based Image Editing with Visual Reasoning
ThinkDrag uses visual reasoning so drag-based image edits follow semantic changes, not only geometry.
Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching
FreeMatching improves identity-preserving dense correspondence for image edits that break physical continuity.
VibeEdit: Image Editing with Canvas Instructions
VibeEdit edits images from on-canvas marks, beating FireRed 79.9 to 67.4 on a 419-case benchmark.
Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching
FreeMatching matches dense correspondences when image edits break classical motion and geometry priors.
UniSlider: Perceptually Uniform Sliders for Continuous Image Editing
UniSlider remaps generative edit strength so perceptual change grows linearly, beating prior sliders on a 300-edit benchmark.
Ideogram says its new model can edit part of an image without messing up the rest
Ideogram 4.5 claims targeted image edits that leave the rest unchanged, now on its platform and API.
Humans are reviewing Copilot users’ bizarre and abusive image-editing requests
Contractors reviewing Microsoft Copilot image-editing requests see uncensored private photos, including sexual and abusive content, with no safety-flagging role.
lilylilith/QI_2.1_AnyAngle — new model trending #30 on Hugging Face
Community LoRA QI_2.1_AnyAngle transfers arbitrary camera angles onto images with Qwen Image 2.1 while preserving style.
20
SolveEdit: Benchmarking Visual Problem Solving in Generative Models
SolveEdit benchmarks goal-driven visual edits; SolveEdit-PLAN lifts GPT-Image-2 from 57.0% to 71.6%.