TexSliders: Diffusion-Based Texture Editing in CLIP Space

TL;DR

This paper introduces TexSliders, a diffusion-based method for texture editing in CLIP space, enabling intuitive, natural language-driven texture modifications without requiring annotated data.

Contribution

It presents a novel approach that manipulates CLIP image embeddings for texture editing, overcoming limitations of existing attention map methods and enabling flexible, identity-preserving edits.

Findings

Effective texture editing with natural language prompts

Preserves identity and minimizes attribute entanglement

No ground-truth annotations needed for the process

Abstract

Generative models have enabled intuitive image creation and manipulation using natural language. In particular, diffusion models have recently shown remarkable results for natural image editing. In this work, we propose to apply diffusion techniques to edit textures, a specific class of images that are an essential part of 3D content creation pipelines. We analyze existing editing methods and show that they are not directly applicable to textures, since their common underlying approach, manipulating attention maps, is unsuitable for the texture domain. To address this, we propose a novel approach that instead manipulates CLIP image embeddings to condition the diffusion generation. We define editing directions using simple text prompts (e.g., "aged wood" to "new wood") and map these to CLIP image embedding space using a texture prior, with a sampling-based approach that gives us identity-preserving directions in CLIP space. To further improve identity preservation, we project these directions to a CLIP subspace that minimizes identity variations resulting from entangled texture attributes. Our editing pipeline facilitates the creation of arbitrary sliders using natural language prompts only, with no ground-truth annotated data necessary.