3D Editing¶
3D Generation
Open Source¶
-
3DGS-Drag: Dragging Gaussians for Intuitive Point-Based 3D Editing Discovered by AI
2026
A point-based 3D editing framework that enables intuitive drag manipulation of real 3D scenes represented by 3D Gaussian Splatting, bridging deformation-based and 2D-editing-based 3D editing.
3d-editingdrag-editinggaussian-splatting -
Alchemy3D: Scaling Versatile 3D Assets Editing with a Million-Scale Dataset Discovered by AI
2026
A unified framework for training and evaluating versatile 3D asset editors, featuring a million-scale dataset for source-aware 3D editing. It addresses the scarcity of high-quality paired training data and provides a robust benchmark for controllable 3D asset modification.
3d-editingasset-editingdataset -
CoIn Discovered by AI
2026
A 3D scene inpainting framework that bridges 2D inpainting models and 3D Gaussian Splatting through a multi-stage consistency pipeline, enabling high-quality object removal and scene completion.
3dgsinpaintingscene-editing -
DesignAgent3D Discovered by AI
2026
An agentic framework for interactive object-level indoor 3D scene editing that couples user intent clarification with geometry-aware execution to resolve ambiguous user instructions and prevent object localization drift.
3d-editingindoor-scenemultimodal-agentscene-editing -
EditVerse3D Discovered by AI
2026
A 3D object editing framework that enables high-quality local modifications under coarse region-of-interest guidance, bypassing the need for precise 3D masks or fully edited 2D images.
3d-editinglocal-editingregion-aware -
FluSplat: Sparse-View 3D Editing without Test-Time Optimization Discovered by AI
2026
FluSplat is a feed-forward framework for cross-view consistent 3D scene editing from sparse views, decoupling the editing process from 3D reconstruction to eliminate the need for test-time optimization.
3d-editingfeed-forwardgaussian-splattingsparse-view -
GeM-NR: Geometry-Aware Multi-View Editing for Nonrigid Scene Changes Discovered by AI
2026
GeM-NR is a fast, training-free pipeline for general multi-view consistent image and scene editing that supports nonrigid changes. It estimates the geometry of the edited scene, projects the edited point cloud onto new views, and uses warping-based conditioning to maintain multi-view consistency even during drastic structural modifications.
3d-editinggeometry-awaremulti-viewnonrigid-editing -
Godot MCP Discovered by AI
2026
Godot MCP integrates AI assistants into the Godot editor, enabling automated scene editing, input injection, deterministic playtesting, and live game state inspection.
ai-assistantgodotmcpscene-editing -
Look Before You Edit: Attention-Guided Camera Placement and Multi-View Alignment for 3D Gaussian Splatting Editing Discovered by AI
2026
LB-Edit is a text-driven 3D Gaussian Splatting editing framework that introduces attention-guided camera placement and multi-view alignment to ensure spatial coverage and cross-view consistency during localized edits.
3d-editinggaussian-splattingmulti-view-alignment -
Omni-3DEdit Discovered by AI
2026
A unified, feed-forward model for multi-view consistent 3D editing that generalizes various editing tasks implicitly. Built upon the pre-trained SEVA model and utilizing a dual-stream LoRA module, it generates edited novel views across tasks like object removal, addition, and color change in a single forward pass without iterative 3D optimization.
3d-editingfeed-forwardloramulti-view -
PriorEdit3D: Learning 3D Editing without Paired Supervision via Generative Prior Distillation Discovered by AI
2026
A feed-forward 3D editing framework that learns instruction-guided 3D editing without paired 3D supervision. It distills 2D visual and VLM-based semantic priors into a 3D-aware model, utilizing a distribution matching regularization to prevent geometric collapse and maintain multi-view consistency.
3d-editingdistillationfeed-forward -
Seed2GS Discovered by AI
2026
A training-free and camera-free framework for extracting target objects from pre-built 3D Gaussian Splatting scenes using a single reference-view grounding, enabling efficient interactive 3D editing.
3d-editinggaussian-splattingobject-extractiontraining-free -
SVGS Discovered by AI
2026
A text-driven 3D object editing framework that uses a single-view image and text instructions to generate multi-view consistent edited images via a cross-domain diffusion model, followed by 3D Gaussian Splatting reconstruction.
3d-editingdiffusiongaussian-splattingsingle-view -
TanGO Discovered by AI
2026
A training-free 3D editing framework that enables adaptive per-token steering in the tangent space of generative dynamics, preventing semantic artifacts and preserving unedited regions.
3d-editingflow-matchingtraining-free -
Thinking in Boxes Discovered by AI
2026
A geometric 3D editing framework that uses 3D bounding boxes as structured specifications to guide large spatial transformations, object motions, and camera changes in real images.
3d-editingbounding-boximage-editing -
VecSet-Edit Discovered by AI
2026
A pipeline that leverages pre-trained Large Reconstruction Models (LRM) to enable effective 3D mesh asset editing from a single reference image.
3d-editinglrmmesh-editing -
ViSculpt Discovered by AI
2026
A training-free, visual-centric agentic geometry editing framework that leverages large language models to perform perception-driven editing of arbitrary existing meshes. It translates creative intent into precise operations while preserving untouched regions.
geometry-editingllm-agentmesh-editing -
WINGS: Reference-Free Gaussian Splatting Inpainting with 3D-Native Generative Priors Discovered by AI
2026
A reference-free 3D Gaussian Splatting inpainting method that operates natively in 3D space, bypassing the need for 2D diffusion models to generate multi-view consistent reference views. It leverages 3D-native generative priors to produce plausible geometry and appearance within masked regions efficiently.
3d-editinggaussian-splattinginpainting -
ARAP-GS: Drag-driven As-Rigid-As-Possible 3D Gaussian Splatting Editing with Diffusion Prior Discovered by AI
2025
ARAP-GS is a drag-driven 3D Gaussian Splatting editing framework based on As-Rigid-As-Possible (ARAP) deformation. It applies ARAP deformation directly to 3D Gaussians, enabling flexible, user-guided geometric transformations and shape adjustments while preserving the original appearance and multi-view consistency.
3d-editingdeformationdrag-to-editgaussian-splatting -
DeMapGS Discovered by AI
2025
A structured Gaussian Splatting framework that jointly optimizes deformable template meshes and surface-attached 2D Gaussian splats. This unified representation overcomes topological inconsistencies and enhances editing flexibility, supporting high-fidelity diffuse, normal, and displacement map extraction.
3d-editinggaussian-splattingmesh-deformation -
2025
Provides drag-style interactive editing of 3D Gaussian splatting scenes, propagating user-specified point drags into consistent 3D geometry changes.
drageditinggaussian-splatting -
Edit3r Discovered by AI
2025
A feed-forward framework that reconstructs and edits a 3D scene in a single pass from sparse unposed images and a text prompt. Instead of optimizing a scene-specific representation, it directly predicts instruction-aligned 3D Gaussians and resolves cross-view conflicts introduced by stochastic 2D editors.
3d-editingfeed-forwardgaussian-splatting -
Free-Form Scene Editor Discovered by AI
2025
A 3D-aware autoregressive framework designed to enable intuitive, physically-consistent object editing directly on real-world images. It supports multi-round object manipulation, allowing users to move, rotate, and scale objects like in a 3D engine.
3d-awarediffusionscene-editing -
Instruct-4DGS Discovered by AI
2025
An efficient dynamic scene editing method that achieves high scalability in the temporal dimension by separating static and dynamic components within 4D Gaussian Splatting.
3d-editing4d-gaussian-splattingdynamic-scene -
MCP for Unity Discovered by AI
2025
An open-source Model Context Protocol (MCP) server that bridges AI assistants with the Unity Editor, enabling LLMs to manage assets, control scenes, edit scripts, and automate workflows directly inside Unity.
automationmcpscene-editingunity -
NANO3D: A Training-Free Approach for Efficient 3D Editing Without Masks Discovered by AI
2025
A training-free framework for precise and coherent 3D object editing without masks, integrating FlowEdit into TRELLIS to edit 3D assets directly in latent space.
3d-editinglatent-spacetraining-freetrellis -
RoMaP: Robust 3D-Masked Part-level Editing in 3D Gaussian Splatting with Regularized Score Distillation Sampling Discovered by AI
2025
RoMaP is a local 3D Gaussian Splatting editing framework that enables precise, text-guided, part-level modifications by integrating a geometry-aware 3D mask generator with a regularized Score Distillation Sampling (SDS) objective.
3d-editinggaussian-splattingpart-level-editingsds -
Tinker: Diffusion's Gift to 3D--Multi-View Consistent Editing From Sparse Inputs without Per-Scene Optimization Discovered by AI
2025
A versatile framework for high-fidelity 3D editing that operates in both one-shot and few-shot regimes without any per-scene optimization, leveraging pretrained diffusion models to deliver multi-view consistent edits.
3d-editingdiffusionmulti-view-consistencytraining-free -
VoxHammer: Training-Free Precise and Coherent 3D Editing in Native 3D Space Discovered by AI
2025
A training-free 3D local editing framework that operates in native 3D space (voxel/latent space) rather than relying on multi-view 2D image editing and reconstruction, ensuring precise preservation of unedited regions and multi-view coherence.
3d-editinglocal-editingtraining-freevoxel -
3DitScene Discovered by AI
2024
A unified scene editing framework leveraging language-guided disentangled Gaussian Splatting that enables seamless editing from 2D to 3D at both global and individual object levels.
3d-editinggaussian-splattingscene-editing -
3DSceneEditor Discovered by AI
2024
A 3D-only editing paradigm for Gaussian Splatting that enables direct manipulation of 3D Gaussians for fast, precise, and controllable scene editing based on text prompts, bypassing the slow multi-step 2D-to-3D projection process.
3d-editinggaussian-splattingscene-editing -
Chat-Edit-3D Discovered by AI
2024
An interactive 3D scene editing framework that leverages large language models to parse flexible user text prompts and coordinate multiple 2D/3D models for complex editing tasks.
3d-editinginteractivellm -
DGE: Direct Gaussian 3D Editing by Consistent Multi-view Editing Discovered by AI
2024
A two-stage 3D editing framework that adapts a 2D image editor to be multi-view consistent, followed by a fast 3D fitting stage on 3D Gaussian Splatting, enabling efficient and high-quality text-driven 3D scene editing.
3d-editinggaussian-splattingtext-driven -
Diffusion-Based Attention Warping for Consistent 3D Scene Editing NEW Discovered by AI
2024
A 3D scene editing framework that warps 2D diffusion attention features across viewpoints using depth estimates from Gaussian splatting to achieve multi-view consistent text-driven edits.
3d-editingattention-warpinggaussian-splattingscene-editing -
DreamCatalyst Discovered by AI
2024
A fast and high-quality 3D editing framework based on Score Distillation Sampling (SDS). It introduces a mechanism to control editability and identity preservation, resolving conflicts with the sampling dynamics of diffusion models to significantly reduce training time while improving visual quality.
3d-editingscore-distillationsds -
EditSplat Discovered by AI
2024
A text-driven 3D scene editing framework that integrates multi-view fusion guidance and attention-guided optimization to achieve view-consistent edits on 3D Gaussian Splatting.
3d-editinggaussian-splattingmulti-view-consistency -
2024
Performs text-driven editing of 3D Gaussian splatting scenes with depth-conditioned and attention-based multi-view consistent image edits.
editinggaussian-splattingmulti-view-consistent -
GSEdit Discovered by AI
2024
An efficient text-guided pipeline for editing the style and appearance of 3D objects represented by Gaussian Splatting. It optimizes the 3D model while progressively varying the viewpoints and applying Score Distillation Sampling (SDS) loss to ensure multi-view consistency.
3dgssds-losstext-guided -
GSEditPro: 3D Gaussian Splatting Editing with Attention-based Progressive Localization Discovered by AI
2024
GSEditPro is a text-driven 3D scene editing framework based on 3D Gaussian Splatting that integrates attention-based progressive localization and Score Distillation Sampling. It allows users to perform precise, localized geometric and appearance modifications on 3D scenes using text prompts.
3d-editinggaussian-splattinglocalizationtext-to-3d -
Instant3dit: Multiview Inpainting for Fast Editing of 3D Objects Discovered by AI
2024
A generative technique that casts 3D editing as a multiview image inpainting task, enabling fast, localized edits of meshes, NeRFs, or Gaussian Splats in seconds without requiring score distillation sampling (SDS) optimization.
3d-editinggaussian-splattinginpaintingmultiview -
Instruct-GS2GS Discovered by AI
2024
A method for instruction-based editing of 3D Gaussian Splatting scenes using a 2D diffusion model to iteratively update rendered images while optimizing the underlying 3D representation.
3d-editinggaussian-splattinginstruct-pix2pix -
LichtFeld Studio Discovered by AI
2024
An open-source native workstation and application for 3D Gaussian Splatting that integrates training, real-time inspection, editing, automation, and export locally on the GPU.
3d-editinggaussian-splattingsoftwareviewer -
MVInpainter: Learning Multi-View Consistent Inpainting to Bridge 2D and 3D Editing Discovered by AI
2024
Formulates 3D editing as a multi-view 2D inpainting task, training a model to generate multi-view consistent inpainted images that can be seamlessly integrated into 3D reconstruction pipelines.
3d-editing3d-reconstructioninpaintingmulti-view-consistency -
splatviz Discovered by AI
2024
A full Python interactive 3D Gaussian Splatting viewer and editor that enables real-time editing, live tracking of training runs, and side-by-side scene comparison.
3dgseditingviewer -
SuperSplat Discovered by AI
2024
An open-source, browser-native 3D Gaussian Splat editor built on PlayCanvas. It allows users to inspect, edit, crop, optimize, and export 3D Gaussian Splat scenes directly in the browser.
3dgseditorgaussian-splattingwebgl -
Texture-GS: Disentangling the Geometry and Texture for 3D Gaussian Splatting Editing Discovered by AI
2024
Disentangles the appearance and geometry of 3D Gaussian Splatting by mapping the appearance to a 2D texture map, enabling high-fidelity texture editing, swapping, and painting on 3DGS models.
3d-editingdisentanglementgaussian-splattingtexture-mapping -
TIP-Editor Discovered by AI
2024
A 3D scene editing framework that accepts both text and image prompts alongside a 3D bounding box to specify the editing region. It enables precise control over the appearance and location of edited objects by utilizing a localized editing pipeline.
3d-editingimage-prompttext-to-3d -
TrAME: Trajectory-Anchored Multi-View Editing for Text-Guided 3D Gaussian Splatting Manipulation Discovered by AI
2024
TrAME introduces a progressive 3D editing strategy for 3D Gaussian Splatting that ensures multi-view consistency using a Trajectory-Anchored Scheme (TAS) and a dual-branch editing mechanism. It tightly couples 2D view editing with 3D updates to prevent error accumulation and maintain high-fidelity visual quality.
3d-editingconsistencygaussian-splattingmulti-view -
VcEdit Discovered by AI
2024
A view-consistent 3D editing framework for 3D Gaussian Splatting that addresses multi-view inconsistency by propagating edits across views using a single-stage diffusion model.
3dgsdiffusionediting -
DreamEditor Discovered by AI
2023
A text-driven 3D scene editing framework that represents scenes as mesh-based neural fields to allow localized editing within specific regions. It automatically identifies target regions based on text prompts and optimizes them using a pretrained text-to-image diffusion model.
mesh-editingneural-fieldstext-to-3d -
Free-Editor Discovered by AI
2023
A training-free, zero-shot text-driven 3D scene editing framework that allows editing a 3D scene by modifying only a single view, significantly reducing processing time.
3d-editingnerftraining-freezero-shot -
2023
Augments 3D Gaussians with identity features to jointly reconstruct and segment open-world scenes, supporting editing operations like removal and inpainting.
editinggaussian-splattingsegmentation -
GaussianEditor Discovered by AI
2023
An innovative and efficient 3D editing algorithm based on 3D Gaussian Splatting that enhances precision and control through Gaussian semantic tracing.
3d-editinggaussian-splattingtext-to-3d -
2023
Enables fast, controllable text-guided editing of 3D Gaussian splatting scenes through semantic tracing and constrained Gaussian updates.
editinggaussian-splattingtext-guided -
2023
Edits reconstructed NeRF scenes from natural language instructions by iteratively updating input images with an image-conditioned diffusion editor.
editinginstructionnerf -
2023
A part-aware NeRF generative model that represents objects as a set of locally defined neural radiance fields enabling part-level control and editing.
editingnerfpart-based -
SC-GS: Sparse-Controlled Gaussian Splatting for Editable Dynamic Scenes Discovered by AI
2023
A dynamic scene representation that decomposes motion and appearance using sparse control points, enabling both high-fidelity dynamic view synthesis and interactive motion editing of 3D Gaussian Splats.
3d-editingdynamic-scenesgaussian-splattingmotion-editing -
2022
An implicit shape representation decomposed into editable Gaussian parts, enabling part-level manipulation and editing of generated neural shapes.
editingimplicitpart-based
