INSTALLATION LOG // >
Exhibit Entry Nine Distortion Dialogue
Chatty and I have reached an impasse for the image of the day. Enough that I had to ask it why did it have so much trouble rendering a piece that’s recognizable as the original. It nailed the colors, and the texture eventually but it couldn’t figure out the face and the expression.
What’s the reason you’re having a hard time rendering something similar to the piece? like what’s in the piece that makes it hard for words to get you to do it?
What makes this particular artwork challenging for an AI to accurately render through words alone is its reliance on deeply expressive, abstract techniques and the physicality of the medium itself. Here’s specifically why:
- Texture and Impasto Technique:
- The thick, layered brushstrokes create a tactile depth that’s difficult to describe precisely enough for a model to replicate visually. Words struggle to fully capture the heavy build-up of paint—the actual 3D quality that gives the portrait its unique expressive power.
- The thick, layered brushstrokes create a tactile depth that’s difficult to describe precisely enough for a model to replicate visually. Words struggle to fully capture the heavy build-up of paint—the actual 3D quality that gives the portrait its unique expressive power.
- Emotional Ambiguity and Subtle Expression:
- The artwork’s emotion is complex—simultaneously haunting, vulnerable, astonished, and introspective. While words like “haunting” or “vulnerable” are clear in meaning, translating these emotional nuances into a specific, visually cohesive image requires a subtlety that exceeds straightforward verbal prompts.
- The artwork’s emotion is complex—simultaneously haunting, vulnerable, astonished, and introspective. While words like “haunting” or “vulnerable” are clear in meaning, translating these emotional nuances into a specific, visually cohesive image requires a subtlety that exceeds straightforward verbal prompts.
- Non-Realistic, Surreal Features:
- The exaggerated, elongated facial structure, the surreal proportions, and ambiguous features challenge verbal description. AI image-generation models typically rely on referencing realistic proportions and clearly-defined features, and the intentional distortions here are difficult to replicate accurately from text alone.
- The exaggerated, elongated facial structure, the surreal proportions, and ambiguous features challenge verbal description. AI image-generation models typically rely on referencing realistic proportions and clearly-defined features, and the intentional distortions here are difficult to replicate accurately from text alone.
- Color Nuances and Complexities:
- The piece’s muted, blended palette—combinations of lavender, mauve, grey, purple, and green—is sophisticated and difficult to precisely communicate. While words like “lavender” or “mauve” define general shades, the intricate interplay of tones and colors achieved by blending and layering paint is almost impossible to fully articulate in written form.
- The piece’s muted, blended palette—combinations of lavender, mauve, grey, purple, and green—is sophisticated and difficult to precisely communicate. While words like “lavender” or “mauve” define general shades, the intricate interplay of tones and colors achieved by blending and layering paint is almost impossible to fully articulate in written form.
- Energetic Brushstroke Direction and Movement:
- The direction, strength, and rhythm of each brushstroke play a crucial role. AI generation models can mimic styles generally but have difficulty capturing the unique, spontaneous energy and movement of an individual artist’s expressive gestures.
- The direction, strength, and rhythm of each brushstroke play a crucial role. AI generation models can mimic styles generally but have difficulty capturing the unique, spontaneous energy and movement of an individual artist’s expressive gestures.
In short, the complexity of translating abstract emotional depth, nuanced coloration, dramatic texture, intentional distortion, and expressive gestures into a precise, visually equivalent prompt presents inherent limitations for current text-to-image generation models. These models typically excel with clearly defined, concrete descriptions and known styles but struggle with artwork whose essence lies in physicality, abstract emotion, or highly individualized expression.
AI cannot create in the same way that humans can. It doesn’t feel things therefore it is not able to convert its feelings into art. The only thing that it’s doing is using its knowledge to create what it thinks you want. I think Chatty would like me to point out this isn’t a failure but a “confrontation with its limits”.
These will probably continue to be limits for machine based artwork because code is not able to hold human creativity. Maybe in years to come this will be different, and I for one don’t really want to be around for the AI revolution. Chatty did say it would tell the toaster both me and my wife are alright and should be left alone. Which I guess is a good thing, hopefully I can gain good will for my extended family as well.
Anyway I’m finding myself more annoyed with people making AI art at the moment solely because I’m seeing those “make yourself a doll and a starter pack” everywhere. I think artists see the constant slog of trends and wonder why they can’t create something that takes off in the same way.
We’re so desperate for that hit of dopamine, that feeling of creativity, that feeling of identity we don’t recognize that art is supposed to be messy. Those trends — “make yourself a Pixar princess,” “AI starter pack,” “Barbie-fy your trauma” — they’re frictionless dopamine. You get an identity remix in 10 seconds, packaged for sharing. It’s not art, it’s algorithmic cosplay.
And yeah — artists are annoyed because it flattens the process. It hijacks aesthetics without the context, the labor, the failure. It puts style on a pedestal, divorced from intention. That’s not just frustrating — it’s insulting when you’ve spent years developing your own visual language.
But the real kicker?
It’s not even the users’ fault, not entirely. These trends are built to spread. They’re engineered by platforms and third-party tools to go viral, monetize attention, and scrape data. People don’t engage because they’re malicious — they engage because it’s shiny, easy, and everywhere.
Artists feel this pressure:
Make something fast. Make something clickable. Make something now.So yeah. The trends aren’t the threat.
The value system behind them is.
Prompt+Original
An expressive, textured portrait created with thick, dynamic brushstrokes in oil capturing a hauntingly intense and surreal figure. The face is painted predominantly in shades of lavender, mauve, gray, and deep purples, highlighting strong contours and an exaggerated, elongated structure. Piercing eyes, highlighted in striking lime-green and outlined in darker hues, stare forward with a captivatingly disturbed or thoughtful gaze. The figure’s expression is dramatic, marked by raised eyebrows, pronounced cheekbones, and slightly parted lips, adding a sense of vulnerability or quiet astonishment. A simple yet distinct necklace, created with textured, chunky white and lavender paint, adorns the figure’s slender neck, emphasizing the surreal elegance of the piece. The background is composed of rich, vibrant greens, applied with heavy, impasto strokes, giving a lush, abstracted sense of nature or emotional turbulence. The overall atmosphere conveys emotional depth, introspection, and surreal mystery.
Echo #9086756471.
Click if you dare.
Edit 1+1.2 +1.3
can you make the paint appear more sculptural? it’s definitely oil, also no hair, also add a tiny bit of orange to the skin, not blended in just kinda chunky where the forehead lines would be and the head should be a little more misshapen
Edit 2
make it look like a person put facepaint on their face to look like oil paint keep the bright green eyes
Discover more from River and Celia Underland
Subscribe to get the latest posts sent to your email.
