Emu Edit

Emu Edit is an image editing model that performs multiple editing tasks based on text instructions. Rather than requiring separate tools for different editing jobs, it handles region-based editing, free-form adjustments, object detection, and segmentation in a single system. The model uses task embeddings and few-shot learning, meaning it can adapt to new editing tasks with relatively few examples. It's trained on a diverse range of editing scenarios, which allows it to handle tasks like changing backgrounds, adding objects, and modifying image regions. The tool performs well across seven benchmarked editing tasks, making it suitable for designers, content creators, and developers who want instruction-based image editing without managing multiple specialised tools.