Skip to main content
GameDev.net gamedev.net

Image Editing News

The latest Image Editing coverage curated for game developers.

Reference-free Human-Object Interaction Editing

A new image-editing framework, InteractEdit, tackles human-object interaction changes without needing a reference image. It aims to preserve subject and object identity while swapping the interaction itself, a harder problem …

Research arXiv cs.GR · 1 month ago
ResEdit: Residual embeddings for precise generative image editing

A new diffusion-editing method adds a residual image embedding to improve both identity preservation and edit precision. For teams experimenting with generative tools, the interesting part is that it reduces …

Research arXiv cs.GR · 3 months, 1 week ago
SeamEdit: A Black-Box VLM-Agnostic Pipeline for Large-Image Semantic Editing

For teams doing large-image editing, the interesting bit is a training-free pipeline that lets closed VLMs act as black-box inpainting engines without retraining. SeamEdit tackles the usual tiled-edit problems—semantic drift, …

Research arXiv cs.GR · 3 months, 2 weeks ago
On the Controllability-Fidelity Frontier in Diffusion Editing

For teams using AI-assisted image editing, this paper is less about a new tool than about the tradeoffs underneath them. It maps the controllability-fidelity frontier in diffusion editing, showing where …

Research arXiv cs.GR · 3 months, 2 weeks ago
Semantic-Structural Alignment for Generative Pictorial Charts

A new diffusion-based system can generate pictorial charts that keep the underlying data layout intact while borrowing the look of reference art. For teams building data-heavy UI, marketing visuals, or …

Research arXiv cs.GR · 3 months, 2 weeks ago
PaintBench: Deterministic Evaluation of Precise Visual Editing

For teams building image-editing or in-game content tools, the big takeaway is that precise visual edits are still hard: across 11 models, the best result in PaintBench was only 17.1% …

Research arXiv cs.GR · 3 months, 3 weeks ago
Materialist: Physically Based Editing Using Single-Image Inverse Rendering

Materialist introduces a novel approach to image editing that leverages neural networks for initial material predictions, optimizing them through physics-based rendering. This advancement is particularly relevant for graphics programmers and …

Research arXiv cs.GR · 4 months, 3 weeks ago
HeadRouter: A Training-free Image Editing Framework for MM-DiTs by Adaptively Routing Attention Heads

The introduction of HeadRouter presents a significant advancement for graphics programmers and artists, offering a training-free image editing framework. By adaptively routing attention heads in multi-modal diffusion transformers (MM-DiTs), this …

Research ACM Graphics · 5 months ago
IMAGAgent: Orchestrating Multi-Turn Image Editing via Constraint-Aware Planning and Reflection

IMAGAgent introduces a significant advancement in multi-turn image editing by utilizing a closed-loop mechanism that enhances context-awareness. This innovation is particularly beneficial for artists and designers, as it reduces errors …

Research arXiv cs.GR · 5 months, 3 weeks ago
Workflow-Aware Structured Layer Decomposition for Illustration Production

A new framework for structured layer decomposition in anime illustration production could significantly enhance the workflow for artists. By breaking down images into semantically meaningful layers, it allows for improved …

Research arXiv cs.GR · 6 months, 1 week ago
Spanning the Visual Analogy Space with a Weight Basis of LoRAs

A new paper argues that one LoRA is too cramped to cover the full range of visual analogies, so it builds a learnable basis of LoRAs and mixes them per …

Research arXiv cs.GR · 7 months, 1 week ago
Alterbute: Editing Intrinsic Attributes of Objects in Images

A diffusion-based image editor can now change an object’s color, texture, material, and even shape while trying to keep its identity and scene context intact. The interesting bit for game …

Research arXiv cs.GR · 8 months, 1 week ago
LooseRoPE: Content-aware Attention Manipulation for Semantic Harmonization

LooseRoPE introduces a novel approach to image editing that allows for precise control over object integration without the need for textual prompts. This method is particularly relevant for graphics programmers …

Research arXiv cs.GR · 8 months, 2 weeks ago
CoatFusion: Controllable Material Coating in Images

CoatFusion introduces a new image editing technique that allows developers to apply controllable material coatings to 3D objects. With a dataset of 110K images, this method preserves the underlying geometry, …

Research arXiv cs.GR · 9 months, 3 weeks ago