CVApr 5, 2022

Text2LIVE: Text-Driven Layered Image and Video Editing

arXiv:2204.02491v2395 citationsh-index: 33
AI Analysis

This enables users to manipulate object textures or add visual effects in images and videos using text prompts, though it is incremental as it builds on existing CLIP-based editing methods.

The paper tackles zero-shot, text-driven appearance editing in images and videos by generating an edit layer composited over the original input, achieving localized, semantic edits without pre-trained generators or user masks.

We present a method for zero-shot, text-driven appearance manipulation in natural images and videos. Given an input image or video and a target text prompt, our goal is to edit the appearance of existing objects (e.g., object's texture) or augment the scene with visual effects (e.g., smoke, fire) in a semantically meaningful manner. We train a generator using an internal dataset of training examples, extracted from a single input (image or video and target text prompt), while leveraging an external pre-trained CLIP model to establish our losses. Rather than directly generating the edited output, our key idea is to generate an edit layer (color+opacity) that is composited over the original input. This allows us to constrain the generation process and maintain high fidelity to the original input via novel text-driven losses that are applied directly to the edit layer. Our method neither relies on a pre-trained generator nor requires user-provided edit masks. We demonstrate localized, semantic edits on high-resolution natural images and videos across a variety of objects and scenes.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes