What is Emu Edit?
Emu Edit is an instruction-based image editing model developed by AI at Meta. Emu Edit pairs an input image with a written request and generates an edited image. The research combines editing and visual recognition in one model. Its demonstrations, paper, and benchmark give image editing researchers material for examining how a multi-task model handles different kinds of instructions.
What Emu Edit can do
Emu Edit handles changes confined to a region as well as free-form changes that can affect the whole image. The public benchmark shows the range of edits used for evaluation, from adding or removing an object to changing the style of a scene. The model’s training also includes recognition tasks such as detection and segmentation.
- Emu Edit can replace an image background.
- Emu Edit can add or remove objects.
- Emu Edit can apply a different style to an image.
- Emu Edit can make localized changes.
- Emu Edit can alter colors and textures.
How Emu Edit works
Emu Edit uses learned task embeddings to guide generation toward the intended task. The training approach puts editing and computer vision tasks into a common generative format. For adaptation to an unseen task, the researchers use a method called task inversion: they keep the model weights frozen and update only a task embedding. The page illustrates this approach with examples of marking drinks and increasing image resolution.
Who Emu Edit is for
Emu Edit’s research materials are relevant to researchers comparing instruction-based image editors or studying how a model adapts to new visual tasks. The benchmark separates different kinds of edits, so researchers can compare outputs on tasks such as changing a background or removing an object. The adaptation experiments also address settings where labeled examples or computing resources are limited.
- Compare model outputs on the public benchmark.
- Examine localized edits separately from whole-image changes.
- Study few-shot adaptation for super-resolution or contour detection.
Emu Edit pricing and research access
Emu Edit’s project page does not state a price, list subscription plans, or describe a free tier for the model. The page links to the research paper and offers downloads of the public benchmark and the images Emu Edit generated on that dataset. Those downloads provide materials for evaluating the published results; the page does not say how to obtain or run the model itself.
Emu Edit strengths and limitations
Emu Edit’s research reports that multi-task training and learned task embeddings improve how accurately the model follows edit instructions. The benchmark outputs make comparisons possible across several kinds of edits, not just selected demonstrations. The reported few-shot results cover examples such as super-resolution and contour detection. The page does not establish how well the same adaptation method would work for every unseen task.
Emu Edit FAQ
What is Emu Edit?
Emu Edit is a multi-task image editing model developed by AI at Meta. Emu Edit follows written instructions to generate edits to an input image. Its training covers region-based and free-form editing as well as computer vision tasks such as detection and segmentation.
What kinds of edits can Emu Edit make?
Emu Edit can change backgrounds, alter styles, add or remove objects, and modify individual regions. Its benchmark also covers changes to an image’s overall appearance and edits to colors or textures.
Can Emu Edit adapt to new image tasks?
Emu Edit can adapt to new tasks through few-shot task inversion, according to the reported experiments. The method keeps the model weights frozen and updates only a task embedding. The page names super-resolution and contour detection as examples.
What is included in the Emu Edit benchmark?
Emu Edit provides a public benchmark covering background changes, whole-image changes, style changes, object removal, object addition, localized edits, and color or texture changes. The project also releases Emu Edit’s generated outputs on the dataset so researchers can compare results.
Is Emu Edit free to use?
Emu Edit’s project page does not state whether the model is free to use or how to access it. The page provides editing demonstrations, a link to the paper, and downloads of the benchmark and the model’s benchmark outputs.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




