DistantNews
Support us
🇰🇷 South Korea /Technology

AI Photo Editing: Partnering with Kim Sung-joo's AI Photo Recipes

From Hankyoreh · () Korean

Translated from Korean and summarized by DistantNews. Read the original for the full story.

At a glance

In-depth Sources not specified Context piece
  • Photographer Kim Sung-joo experimented with ChatGPT for advanced photo editing tasks, including analysis, correction, and generative AI.
  • The AI demonstrated an ability to identify key elements, adjust exposure, color, and composition, and even remove objects like utility poles.
  • While AI offers powerful editing capabilities, the article suggests that detailed instructions and training are crucial for achieving desired results, positioning AI as a potential partner.

In an era where images often replace words, the art of photo editing is gaining traction. Photographer Kim Sung-joo explored the potential of artificial intelligence by tasking ChatGPT with editing photographs. He assigned the AI the role of a skilled photographer, instructing it through three stages: preparatory analysis, correction and enhancement, and a post-editing report.

The AI demonstrated an ability to identify key elements, adjust exposure, color, and composition, and even remove objects like utility poles.

— Kim Sung-jooDescribing the AI's editing capabilities on a snowy mountain landscape.

Using the GPT-4 model, Kim began with a landscape photo of a snowy mountain. The AI identified the sky and snowy mountains as strong points but noted distracting elements, a color imbalance towards teal, and weak contrast. The resulting edit maintained the original feel while enhancing clarity and depth. Kim noted that the AI's ability to adjust aspect ratio and remove objects like utility poles, and even use generative AI to emphasize perspective, exceeded his expectations.

The AI's ability to adjust aspect ratio and remove objects like utility poles, and even use generative AI to emphasize perspective, exceeded his expectations.

— Kim Sung-jooReflecting on the AI's performance in editing a snowy mountain landscape.

When editing a photo of the Manhattan skyline, the AI focused on the sunset, adjusting saturation and creating a delicate gradient between pink and blue hues. It also improved the composition by reducing the sky's area and straightening tilted buildings. However, the AI's interpretation differed from Kim's intent, which was to highlight the iconic New Yorker Building. Upon receiving this feedback, the AI adjusted the image to meet Kim's specific request, even naturally restoring image quality degraded during enlargement.

Upon receiving this feedback, the AI adjusted the image to meet Kim's specific request, even naturally restoring image quality degraded during enlargement.

— Kim Sung-jooDetailing how the AI responded to specific instructions for editing the Manhattan skyline photo.

Through these experiments, Kim realized that AI-assisted photo editing is not a simple process. It requires detailed setup and instructions to align with user needs. While AI can reconstruct images beyond the capabilities of standard editing tools, it necessitates a significant learning process. Kim plans to continue refining his use of AI, viewing it as a valuable partner for future creative endeavors.

AI-assisted photo editing is not a simple process. It requires detailed setup and instructions to align with user needs.

— Kim Sung-jooConcluding thoughts on the complexity and requirements of using AI for photo editing.
About this summary

Originally published by Hankyoreh in Korean. Translated, summarized, and contextualized automatically by DistantNews, with a note on how the source frames the story. Not individually reviewed before publishing. How this works.

More Stories

From Our Blog