Douyin's AI Image Expansion Goes Viral
-
Recently, AI-powered image expansion has sparked a trend on Douyin. A blogger named @Zhan Meili gained popularity by posting photos generated with AI image expansion technology. In these photos, the AI not only enlarged the original images but also added elements like moons and magical street lamps, attracting nearly 400,000 likes and over 50,000 shares.
AI image expansion quickly climbed the trending charts, with related videos amassing over 10 million views on Douyin. However, as netizens began using AI image expansion for creative purposes, a series of bizarre and amusing works emerged, propelling AI image expansion to viral fame.
In some creative works, netizens used AI image expansion to blend human figures with imaginative scenes. For example, friends posing in fur coats often turned into sheep or dogs, and couple photos expanded with AI frequently included an unexpected third person, leaving the uploaders both amused and frustrated. This creative approach has drawn widespread attention on social media.
However, some works also highlight that AI-generated outpainting images may appear unrealistic in certain scenarios. This could be related to issues in the model training and generation process. Factors such as insufficient training data, overfitting, unclear training objectives, improper hyperparameter selection, noise and artifacts, and challenges in handling edge cases may all contribute to the lack of realism in generated images.
Although AI image outpainting has gained significant attention in a short period, it's important to note its limitations in generating realistic images. With continuous technological advancements and improvements in the future, these issues are expected to be resolved, enabling AI image outpainting to leverage its advantages in broader application scenarios.