The 'Ridiculous AI Image Expansion' Trend Goes Viral, AI's Hilariously Bad Results Are Killing Me
-
Folks, I'm literally dying from laughter at Douyin AI image expansion feature –
It's all about delivering 'unexpected' and 'infuriating' results~
For example, a loving couple's photo before AI expansion had strong devotional vibes:
Source: Douyin @Happy Wildman
However, after some "magical operations" by AI image expansion, the style undergoes a complete 180-degree reversal:
Source: Douyin @Happy Wildman
After seeing the results, netizens were both amused and inspired, joking that the outcomes were "more devout" and "too motivational"
.
Many people originally wanted to use it to enlarge their photos, but the AI expansion results turned out to be absurd.
Consequently, more and more netizens started posting their own masterpieces of AI image expansion, showcasing several more "unexpected" creations.
AI image expansion masterpiece one, titled "Hold Her Foot":
AI image expansion masterpiece two, "It Takes Ten Years to Grow Trees, But a Hundred Years to Cultivate People":
AI image expansion masterpiece, named 'Wishing You Happiness':
To be honest, it feels like watching a grand documentary titled The Legend of AI Image Expansion.
However, how could such a spectacle be complete without the participation of The Legend of Zhen Huan?
To verify whether AI image expansion is truly this bizarre, we conducted a hands-on test—
Material: The Legend of Zhen Huan; Software: Jianying.
From an operational perspective, it's quite simple. Just open the Jianying App, upload your image, click on "Effects", then "Image Play", and select "AI Image Expansion".
Next, enjoy more highlights of AI image expansion...
"Zhen Huan's Journey to the West": AI directly expanded Zhen Huan's background into a Western palace setting
.
Similar examples include this...
Feed a fat orange cat (His Majesty) into the AI, and the image expansion can return us a pregnancy photo...
The group photo of the concubines, under the grand strokes of AI image expansion, instantly dresses them in jeans and sneakers, strolling through streets blending Eastern and Western styles.
There's an even more advanced technique—feeding AI-altered images back into the AI for further expansion.
For example, the pregnant orange cat photo we obtained earlier yields even more 'astonishing' results when expanded by AI:
Yes, legs were added, along with a suitcase. The AI extended the photo's narrative to 'running away while pregnant'...
The stories of the imperial concubines also received bold innovations, transforming from ancient palace dramas to modern settings.
It must be said that AI image expansion has truly given 'The Legend of Zhen Huan' ten thousand possibilities.
...
At this point, some friends might ask: If AI image expansion is so "weak", why do people still use it?
As the saying goes, even AI can make mistakes. What we've shown are just its failed or illogical cases.
Since the launch of this feature, most image expansion results have been remarkably impressive.
For instance, partial wedding photos of newlyweds can be expanded to include broader scenes in a natural and heartwarming way:
Street photography can also benefit from expanded backgrounds when desired.
So the next question is:
When it comes to AI image expansion, Midjourney, Photoshop, DALL-E 2, Stable Diffusion, and others have all introduced similar features, with some similarities in their underlying principles.
Technologies like DALL-E, Stable Diffusion, and Photoshop's Generative Fill utilize a technique called Outpainting.
Outpainting is an image processing technique, opposite to Inpainting (internal image filling), that synthesizes new content harmonizing with the original image's content, style, and context to expand its visual boundaries.
Outpainting typically relies on deep learning models, including content-diffusion-based, GAN-based, and semantic-understanding-based approaches.
Moreover, AI image expansion isn't just about increasing pixel count but more importantly involves enhancing image resolution.
For example, CNNs are commonly used neural networks in AI image expansion. By learning from numerous low-resolution and high-resolution image pairs, they understand image features and learn how to reconstruct high-resolution images from low-resolution ones.
Super-resolution technology enables models to fill in missing pixels in low-resolution images, thereby generating higher-resolution images.
During the image expansion process, attention must be paid to detail enhancement and noise suppression to preserve image quality.
Although significant progress has been made in AI image expansion technology, real-time processing capabilities still need improvement. Generally, higher-quality image expansion requires more computation time.
Many existing AI image expansion tools have significantly improved generation speed, but the logical consistency of the output remains somewhat random in quality.
Even when AI image expansion sometimes produces absurd results, this "abstract style" has brought much amusement to netizens
.Some netizens have commented that the most entertaining part of Douyin (TikTok) recently has been the AI image expansion comment sections.
Because the grand documentary 'AI Image Expansion Legend' is still being updated...
So today, have you AI-expanded your images?