DragGAN is a groundbreaking AI image editing technology. Imagine, with just a few clicks and a little dragging on an image, you can freely change the pose and expression of the people in the picture, adjusting them to your desired angles. This article will introduce the features and characteristics of the revolutionary photo editing software, DragGAN, as well as provide a hands-on review and usage guide for the online trial version.
What is DragGAN AI?
DranGAN is an image editing software developed by the Max Planck Institute. Anyone can use it to deform images by ‘dragging’ control points on the image, manipulating the poses, shapes, expressions, and image layouts of various subjects, such as animals, cars, humans, landscapes, and more, to generate natural and realistic images.
How DragGAN AI Running?
The paper ‘Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifold,’ published in August 2023, unveils the technology behind DragGAN. Its operational mechanism involves two steps. First, the software extracts features from images, which serve as the basis for image transformations. Then, the images go through training on a researcher’s image database. This training model enables the flexible manipulation of images according to desired changes.
4 Features of DragGAN AI
The groundbreaking technology of DragGAN made an immediate impact on the AI-generated image and image editing technology community. It simplified what used to be a time-consuming and manual retouching process. It primarily features the following aspects:
- Precision Control: It allows precise control over images using control points, enabling you to move objects to desired positions and angles.
- Beginner-Friendly: It’s easy to use, making it accessible even to those who haven’t used any photo editing software before.
- Rapid Generation of Realistic Images: Images generated by DragGAN closely resemble real photos. The areas edited do not exhibit the typical distortion seen in conventional photo editing software. Moreover, the generation process typically takes around one minute.
- Open Source: The open-source code for DragGAN can be found on GitHub, making it accessible for anyone to use and modify.
DragGAN Web Demo Testing and Using Guide
Currently, DragGAN’s Github provides an online demo version through Hugging Face. Next, we will showcase the testing results using the online version:
DragGAN Hugging Face Web Demo：https://huggingface.co/spaces/DragGan/DragGan
Select a model and test image
The online demo version currently offers two default training models for selection, one for dogs and one for portraits. Each model comes with a built-in dataset of 1 billion images, and you can adjust the seed to choose the image you want to edit.
Add control points and begin editing
Next, you can click anywhere on the image to add control points, and then simply press ‘Start’ to begin witnessing and enjoying the magical retouching process! You can also optimize the image by selecting the machine learning latent space. Typically, working in the ‘w+’ space results in better generation, while the ‘w’ space is faster for generating.
First, I selected a picture of a small dog. I wanted to reduce the extent to which its mouth is open, so I added two control points on the mouth (as shown in the upper image). The retouching effect of DragGAN is depicted in the lower image.
The degree to which the dog’s mouth has been reduced in the generated image is indeed subtle, and the image appears natural. However, the dog’s face and body have also been reduced in size. If you don’t want areas other than the mouth to be adjusted, you can use the next feature, ‘masks.’
Use masks to limit the transformation area
The main function of masks is to restrict the deformation area of the image. This feature allows for fine-tuning the details of the image, making it very convenient!
First, we click on ‘Edit Flexible Area,’ and a circular brush will appear on the image. Next, use the brush to paint the area you want to deform black. Then, press ‘Add Points’ to add control points on the image, and finally, press ‘Start’ to view the generated result.
As shown in the image above, the generated result after using the mask is noticeably less in terms of the deformation of the dog’s body compared to when no mask was used.
以上就是線上試用版的基本功能展示，不過DragGAN目前還在開發階段，所以沒辦法上傳自己的照片進行修圖，當然，它的功能與潛力遠遠不只以上提到的這些功能，期待這項技術的發展可以早日成熟，並正式成為AI產品，造福廣大的平面設計師們。
The applications of DragGAN AI include:
- Restoring fine details in photos.
- Enhancing facial expressions in both people and animals.
- Changing the posture and clothing length of individuals.
- Adjusting the size and angles of objects.
Recommendations for AI text generators：
In addition to AI drawing generators, AI text generators are now a new form of creative expression. Here at Genape, we offer a variety of AI text templates for your use. Not only can they assist you in your writing, but they can also provide you with endless creative inspiration. Click below to start your creative journey!
