Meta Emu Edit

What is Meta Emu Edit?

Meta Emu Edit is a research tool from Meta’s AI team built for image editing and benchmarking. It applies artificial intelligence to the editing process and is connected to Emu Video, suggesting applications beyond still-image editing. A research paper documenting the model is publicly available.

Who created Meta Emu Edit?

The tool was developed by a team led by Emily Chang. Public detail about the team is limited, but the project takes a research-driven approach to editing software rather than a conventional one.

What is Meta Emu Edit used for?

• Accurate, instruction-based image editing
• Multi-task training across many image editing tasks at once
• Region-based editing that targets specific areas within an image
• Free-form editing
• Computer vision tasks such as object detection and image segmentation
• Task embeddings that help the model interpret editing instructions
• Improved accuracy in following those instructions
• Few-shot learning, adapting quickly from limited data
• Task inversion, which lets it adjust to new tasks
• A benchmark dataset for testing instruction-based image editing models
• Generative tasks that create new content
• Background alterations and image style changes

Who is Meta Emu Edit for?

It suits anyone who edits content regularly: researchers, video editors, journalists, copywriters, academic and technical writers, marketing professionals, and social media managers.

Getting started

Emu Edit is accessible through Meta’s MetaDemosLab website. The Emu Edit section includes a “Get Started” walkthrough of the basics, along with an official user guide and tutorials covering the multi-task model, region-based and free-form editing, and the supported computer vision tasks.