Analysing Images to AI Art with ChatGPT-4 and Midjourney
Last week, I was experimenting with AI, combining computer vision technology and ChatGPT-4's language capabilities to analyze some of my past work and generate textual descriptions. These descriptions were then used as prompts for Midjourney to create AI-generated images that closely resembled my work. The descriptions weren't 100% spot-on, and I chose not to alter the prompts to better match my original work. I plan to rewrite the prompts later and will share them in another post.
I had no idea how this experiment would turn out. The first test and the outcome—both the description from ChatGPT analyzing my image and the images Midjourney produced—left me speechless for a couple of minutes.
This shouldn’t be a surprise, though. ChatGPT can 'see,' 'listen,' 'talk,' analyze, code, and much more. It's just hard to remember this since it wasn't possible to do such things just months ago. It's like you have to develop a new way of thinking, sort of, I think.
Scroll to see the original work, what I prompted, the descriptions and the AI-generated art.
-
I uploaded an image of a pirate to ChatGPT and prompted it with the following request:
'Describe in as much detail as possible what this pirate looks like. Include all details.'
The result of the description from ChatGPT was as follows:
The person in the image is dressed as a classic, stereotypical pirate. He has a full, thick beard and long hair, both of which are a light brown or auburn color. He is wearing a large, black hat with a red band wrapped around it. Over his right eye, there's a black eyepatch, suggesting the iconic pirate image. He has a white, ruffled shirt underneath a red pirate coat with black embroidery and large black buttons. The coat has a worn-out look with deliberate fraying at the edges. Around his neck, there's a dark necklace with a silver ornament or medallion. His right hand is holding what appears to be a flintlock pistol, indicative of the historical period associated with piracy.
His trousers are also red, matching his coat, and tucked into black leather boots with large folded cuffs. The boots look worn and used. The pirate has several items on his belt, including what looks like a silver cup and possibly a sheath for a sword or dagger. His stance is confident and slightly imposing, with his legs spread apart and hands on his hips or holding the pistol.
In the background, there is a beach and the sea with a pirate ship in the distance on the left side. The sky is orange and cloudy, suggesting either sunrise or sunset. There are some seashells and a crab on the sand near his feet, enhancing the seaside theme. The overall tone of the image is dramatic, with an emphasis on adventure and the romanticized pirate life.
The first image is my original Photoshop work, while the rest are from Midjourney, created in various art styles. There is no end to the different art styles you can explore with Midjourney.
