Aykys Salchak
A graphic designer with 4+ years of experience working as a creative, guided by market trends and design expertise. Currently designing at Whalar, Part of Accenture Song, and pursuing an IDM, M.S. degree at NYU Tandon.
A graphic designer with 4+ years of experience working as a creative, guided by market trends and design expertise. Currently designing at Whalar, Part of Accenture Song, and pursuing an IDM, M.S. degree at NYU Tandon.
+ Info
Gen AI As Creative Material
October 17, 2026
This is a beautiful painting by Child Hassam named “The New York Window” painted in 1912. I absolutely loved the use of color and the impressionist brush strokes. I find subject matter to be very fascinating, especially with it pertaining to New York but the overall ambiance of the painting is somber and somewhat serious. Although the colors are bright, I find the woman’s expression to be calm but stern, even though the sun is beautifully shining from the window. In today’s age we associate New York City to be lively and exciting, but from the 1912 perspective, it looks calm and solemn.
Technical Description &
AI Output
Chat GPT Nano Banana Seedream
Experimentation 01: Collage Made in Photoshop
I felt that I missed the mark on describing the background of the painting when initially making the technical description. Therefore, for the next steps to distort/collage, I added in the modern NYC backgoround to juxtapose the painting style and subject from over a hundred years ago. This was my first experimentation so I made only one change.
Techincal Description &
AI Output
In 1912, a woman was sitting right outside her window, her body facing the front of the camera but tilted to the right. She is wearing a blue draped and comfortable dress with her hair pinned back, in a more relaxed style where you can see her brown waves. She is sitting on a wooden chair with a green back and next to her is a bowl of assorted fruit, yellow and red apples, with grapes at the top. Right behind her is a window, with white see-through drapes with red curtains drawn to the side. The curtains and window take up the majority of the background. This is an impressionist painting with bright but also muted colors with a bit of a serious tone, despite the bright colors and the bright sunny day that is implied to be outside. Despite being in 1912, the outside view is of modern New York. The view is of skyscrapers, especially the one world trade center, and with closer buildings having some black and white graffiti.
Experimentation 02: Changing Timeline and Style
I wanted to see what AI would generate if I made it a modern painting with a more realist style. I wanted to make all the elements of 2026, with the clothing, background, and the medium to be a photograph. I also wanted to remove any elements of how the woman looked, to see if the AI would generate the same exact woman. I was particualry interested if the woman generated would be of the same race and hairstyle that was in the original painting. The women in all three generated images had the same hairstyle but there was more variety in race.
Techincal Description &
AI Output
A modern woman in 2026 is sitting right outside her window, her body facing the front of the camera but is tilted to the right. She is wearing modern style clothing, a hoodie and blue jeans. She is sitting on a wooden chair with a green back and next to her is a bowl of assorted fruit, yellow and red apples, with grapes at the top. Right behind her is a window, with white see-through drapes with red curtains drawn to the side. The curtains and window take up majority of the background. This is a photogrpah with bright but also muted colors with a bit of a serious tone, despite the bright colors and the bright sunny day that is implied to be outside. The outside view is of modern New York. The view is of skyscrapers, especially the one world trade center, and with closer buildings having some black and white graffiti.
Chat GPT
Nano Banana
Seedream
Experimentation 03: Changing Art Style
For my next experimentation, I decided to try completely changing the art style and making a Picasso-esque piece with simplified circular shapes. I used the colors from the original painting but heavily brokedown the piece into ovals and circles.
Techincal Description &
AI Output
Circular shapes that are used to depict a woman sitting near a window sill. Her head and hair are depicted as circles. Her body is a big blue oval with the green chair another oval. The apples on the side are also depicted in circles with yellow, red, and brown. The background has rectangles in the back to depict the window sill, the the pinkish yellow light coming from the window, and the red curtain drawn to the side. The overall painting is a simple, cubist, Picasso-esque style.
Design Takeaway
I found these AI tools to be surprisingly good at what they did. I was shocked that just by my technical description and my quick sketch, the AI was able to somewhat closely replicate the original design. Even the style of dress was very similar to the original painting which made me think of whether the AI knew what artwork I was referring to.
I do think the AI models did do a good job of interpreting my experimentations, as well. The first experimentation was a collage with the painting having a modern NYC city skyline background. I enjoyed seeing the interpretations with the 1912 woman in modern day NYC. The abstract paintings from the third experimentation were also successful and generated my idea well from my quick design.
I do think that some AI was better than others. Chat GPT did a great job of interpreting my ideas and even Seedream did a great job, even though I had never heard of this AI tool. Both Chat GPT and Seedream yielded similar results, with my first AI generated images being eerily similar to each other. I think Gemini/ Nano Banana was the weaker AI model. The first generated image was barely impressionist and the second image barely looked like a photograph and the overall layout was off. However in the last experimentation, I enjoyed Nano Banana’s painting the most, with the use of colors and paint strokes to show a cubist painting.
However, I do think that there are major limitations with AI. I think the impressionist style that the AI models tried to replicate was repetitive and looked too perfect to be made by an artist. The beauty of impressionist art is being able to see the artist’s paint strokes and how their mind was able to interpret the impression of various shapes. The AI models also did not account for diversity. In my experimentations, I did not specify the race of the models and although Chat GPT and Seedream tried to include some diversity in their image, the overall facial aesthetics of the women in all three renderings were very similar.
I think AI can be a good tool for helping ideating and bringing quick ideas to life. It helped me see how my quick experiments and written prompts could be interpreted well without wasting much time. However, I would say that this should only be part of the initial ideating process and not as a final result. The outputs of all three of these AI models were too uniform, not diverse, and lacked that human touch of imperfection and thought.