35% of adults now use Artificial Intelligence tools daily, but over half are concerned that their jobs are threatened by AI (UK IPSOS survey, January 2026). Artists too are fearful of replacement by AI, because of its ability for ‘mechanical creativity’ (Anirveda, 2024). Instead of pessimism, artists like Pindar Van Armen argue that successful artist-AI partnership can happen. Following suggestions by Jennifer Aue in Medium in 2018, I decided to use my own artwork in a practical investigation of strengths and weaknesses in AI-enhanced art.
Inspired by early examples like Chris Rodley’s ‘Dinosaur X Flower mash-up’ (2018), I examined how AI ‘sees’ art by asking it to combine paintings of mine from a recent series I created. The first objective was to ask if AI can make that series more cohesive. My second objective came from Aue’s suggestion that AI’s real strength is ‘style transfer’ based on its huge library of information. I asked AI to combine my work with historical and contemporary painting styles, reflecting on the results to see whether AI could create something new by introducing styles I would not normally use.
The Art Works and Process Used
Whose work would you like to see combined with your own? There are endless possibilities. Perhaps I’d like to be an ‘old master’. I included Rembrandt because of his distinctive use of light and shadow. Then who wouldn’t want to be more like Picasso? I asked AI to combine my work with his style, setting a big challenge knowing that Picasso himself used so many styles across his career. Finally, because of the bright, vibrant approach I favour in my work, I asked AI to combine my paintings with Dystopian art for a complete contrast of styles.
I used Firely on Creative Cloud for my experiment. I fed in visual and written prompts, writing ‘Depict this image in the style of …’, ’Mix these images in the style of …’, ‘Create a painting-like image …’, asking it to use Rembrandt, Picasso, and Dystopian in sequence. As to my own work, I first provided two paintings from my series on sea life, which feature ‘Finding Nemo’ and ‘Sponge Bob’ as inspiration. Separately, I used two paintings from a series I created on Graffiti, to help explore how robust the AI experiment results would be.
Less successful styles: Picasso
AI struggled with visualising my work in the style of Picasso. It seems to take its cue from his Cubist phase, but the results were one-dimensional. Apart from breaking up the image in the style of a cracked glass, there was little new that came from rendering my work in another style using AI.
‘TV Tank’, Picasso style, by AI
More successful: Rembrandt, Dystopian
When AI combined my art with Rembrandt and Dystopian styles the results were more effective. The use of Rembrandt was intriguing and amusing, it rendered my artwork in a Seventeenth Century style – depicting the ‘error messages’ in parchment, the colour bars in layers of oil paint, and placed the screen in a wooden frame, as if Rembrandt was watching TV. This mash up combination brought new aspects into the work I wouldn't have thought of.
More successful: Rembrandt, Dystopian
When AI combined my art with Rembrandt and Dystopian styles the results were more effective. The use of Rembrandt was intriguing and amusing, it rendered my artwork in a Seventeenth Century style – depicting the ‘error messages’ in parchment, the colour bars in layers of oil paint, and placed the screen in a wooden frame, as if Rembrandt was watching TV. This mash up combination brought new aspects into the work I wouldn't have thought of.
The AI translation of my work into Dystopian style was interesting as well. Illustrating this with ‘TV Tank’, It added new features, introducing Graffiti and rusting the TV, including text about a “failed simulation”, and glitching the screen itself. As Aue predicted, it was able to apply its library of style and technique to my work, at least using Rembrandt and Dystopian Art.
Mixing my two sea life art works in the styles of Rembrandt and of Dystopian art
The next challenge for AI was to combine two of my pieces of art into one. The added twist I introduced was by asking for this in the style of the same artists as above.
When AI merged ‘TV Tank’ and ‘Best Day Ever’ Rembrandt-style, it again created interesting visual features: The hand reaching in to change the channel (no remote controls in the 17th Century!), a skull placed in the frame, carefully setting the TV on a soft fabric. With great lighting and finer touches AI again responded well in manner and style. But what about combining the two paintings? It did so in a patchwork like slotting jigsaw pieces together. Sponge Bob is planted on the wider image but clearly not integrated into the scene as an artist would do.
'TV Tank', In the style of Dystopian Art, by AI
Similar results emerged when AI combined the same two paintings in Dystopian style. It broke the image up into four squares, again separating them rather than bringing them together. While failing to merge or integrate it nevertheless introduced visually interesting features based on the art style I requested.
Mixing all four images using Picasso, Dystopian, and Rembrandt styles
AI has done some heavy lifting so far, with its strengths in style, lighting and amusing features. Yet is had weaknesses like not integrating paintings and failing to do well in handling Picasso’s style. I then asked for a big finish – asking AI to combine both sea life paintings with my two Graffiti paintings, and requesting this in the styles of Picasso, Dystopia, and Rembrandt.
AI obliged with a Dystopian image that handled everything in a similar way to before. But this time there’s more integration. It portrays all pieces on a ‘canvas’ it created of large street scene. Notice how it portrays the Graffiti image as a backdrop on the wall, with the girl from the Graffiti image and the car interacting well with the scene. AI is consistently good with amusing features – some of my fish, previously healthy, are now bleak skeletons.
Yet AI again failed to work with Picasso’s style. Once more it falls back to breaking up the images into smaller parts. The effect is to create a series of smaller images, without integration.
TV Tank’, ‘Best Day Ever’, ‘Graffiti Number 1’ and ‘Graffiti Number 2’, merged and Dystopian style, by AI
Finally, asking for a merge of my four paintings in the style of Rembrandt brought the best mash up. The brushstrokes and lighting effects create a backdrop that looks like a Rembrandt painting. Within the scene it created, AI makes my Graffiti into a wall hanging portrayed in shadow and shade. The young woman from my Graffiti painting is now in Seventeenth Century pose, with the car and starfish from my paintings set out on the table in front of her. The school of fish is no longer on TV but displayed in a picture frame.
What’s my contribution? Can AI start from scratch
Having seen how AI reacts to my art, I wanted to see whether my own contribution was actually vital. Instead of giving my paintings as prompts, I merely gave AI descriptive word prompts that named components from each of the paintings (‘sports car’, ‘Sponge Bob’, etc.). I asked it to create new images, but no longer having access to my paintings, it had to be more creative in its own right. The results for each art style were superficial and one-dimensional. It was reassuring that my artwork had contributed something original which AI built on, because the generic images that it created without my input were not convincing.
‘TV Tank’, ‘Best Day Ever’, ‘Graffiti Number 1’ and ‘Graffiti Number 2’, merged and Rembrandt style, by AI
Prompts: ‘Create a painting-like image using iconography from Sponge Bob, an old TV screen, graffiti images, a young woman, and a sports car’, Dystopian Art style, by AI
Prompts: ‘Create a painting-like image using iconography from Sponge Bob, an old TV screen, graffiti images, a young woman, and a sports car’, Picasso style, by AI
Prompts: ‘Create a painting-like image using iconography from Sponge Bob, an old TV screen, graffiti images, a young woman, and a sports car’, Rembrandt style, by AI
This experiment began with fears about AI. But inspired by Aue’s confidence that AI “can change what we communicate by virtue of how we create it and who we create it with”, I learned that AI’s mastery of style transfer could add elements that I could use for further inspiration. But was this AI contribution actually creative? AI’s contribution would not have happened without my input, and it was unable to do much without my original paintings. Jerry Saltz maintains that, because AI does not create art from the lens of human experience, it ‘averages out’ second-hand interpretations. But in a recent article on Contemporary Lynx, Alicja Stapór argued that AI is a tool for artist brainstorming before elevating the work to another level.
AI displayed strengths in mimicking artistic styles and introducing unexpected features. It was high performing in light, detail and shadow. But AI was also mechanical, taking a jigsaw approach to combining paintings. The mash up approach of artist and AI is an increasingly common feature of how artists create or modify their work. Specialised platforms like Leonardo.Ai and Layer (and even mainstream apps like Copilot and Photoshop) allow users to transfer, mix and blend styles. My conclusion was that AI is a tool that can move the human creative process forward, and that its adds to human artwork rather than replacing it – for now …