UPDATE!
Since I use ChatGPT on a daily basis, I realize how much I miss it once it's not available. I feel like a kid who misses his favorite toy when it's not available. Who would have thought that? Almost alarming!
GPT-4 is more precise, smarter, and even more context-sensitive, but even here there is room for improvement. Sometimes even when using GPT-4 (with ChatGPT Plus) I stare at a blank screen because the system is overloaded.
PART 1
Can GPT-4 really analyze images?
The new image analysis feature is quite impressive, but only if it works smoothly. Due to the lack of an upload function for images and the need to use publicly available links (web content), it is not yet fully practicable for me in everyday life.
NOW THE TEST: GPT-4 image analysis of my work "Mangrove d'Or", which - just to clarify - was definitely created WITHOUT AI! 😉
Here is the answer from ChatGPT (GPT-4)
"Mangrove d'Or" painting shows an impressive depiction of a golden mangrove landscape. The composition depicts several trees with characteristic roots and branching branches. They form a dense network that dominates the scene. The golden color of the trees and the background create a warm, radiant atmosphere that emphasizes the beauty and richness of these unique ecosystems.
The artist has worked with different shades of gold to create depth and texture in the scene. The use of light and shadow makes the trees appear three-dimensional and gives a sense of space. Amidst the mangroves, fine details such as individual leaves and branches are visible, hinting at the complexity and diversity of these habitats.
The artist's technique is meticulous and precise, emphasizing the detailed depiction of the mangrove landscape. Overall, the painting "Mangrove d'Or" conveys an appreciative and respectful view of these fascinating, yet often overlooked habitats."
Hey, something's missing!
What is clearly missing from ChatGPT's reply to me is any mention of who is the creator of the image. Look at the answer:
PART 2
So how can GPT-4 image analysis be used in Midjourney, and does it really work?
The Midjourney architecture in detail as a guidline:
[Subject], [Adjective], [Setting], [Art style], [Type of art], [Texture], [Visual style], [Presentation format], [Color scheme], [Subculture], [Lighting], [Realism level], [Detail], [Environment], [Software], [Resolution]
CONCLUSION
In this experiment, I used my own creation (Mangrove d'Or) as a base. The idea that now basically anyone can make a similar copy of my work using Midjourney by simply screenshotting it, not only causes discomfort, but is also extremely frustrating and a big hassle. It's painful enough that others imitate or even steal my creations.
Your turn!
What do you think of this analysis by GPT-4? Let me know what you think - share your feedback in the comments!
I look forward to hearing different perspectives and interpretations.