Summary

An artist reflects on their 1990s artwork by using AI to recreate and enhance their old sketches, with mixed and often humorous results.

Abstract

In the 1990s, the author, then an aspiring art student, faced criticism for their artwork, leading to a brief departure from art. Decades later, with the advent of AI image generation, they revisit their old sketches, curious to see how AI platforms like DALL-E, Midjourney, and DreamStudio would interpret and possibly improve them. The author scans and inputs their drawings into these AI systems, resulting in a variety of outcomes that range from unexpected reinterpretations to surprisingly accurate renditions. The article showcases a selection of these AI-generated pieces alongside the original drawings, providing a commentary on the evolution of art and technology, and the subjective nature of artistic interpretation.

Opinions

The author initially believed their artwork was of high quality, only to be met with disapproval from art instructors.
Despite the setback, the author eventually returned to art, rejecting traditional gatekeeping in the art world.
The author finds humor in the AI's interpretations, such as turning a monster into a pet python or altering the concept of a drawing entirely.
There is a note of irony in the AI platforms flagging content for sensitivity, such as cigarettes or nudity, which the author had to censor manually in the original artwork.
The author appreciates the AI's

I Let AI Recreate My Shitty Drawings From The 1990's

The results were not quite as I expected

In 1995, I was a hopeful young buck headed off to art school. I quickly learned that despite my friends and family saying so, I was not the best artist around. Needless to say, at the end of my first year when it was time to review portfolios, I got a rude awakening.

The instructors — who each examined the work — just shook their heads disapprovingly at my sketches and paintings. I left the review room with a broken heart and a red face, soon after taking a job at a camera store.

Anyways, while I eventually started making art again (screw the art gatekeepers), I can acknowledge that my artworks in the 1990’s were… well, not that good. I recently dragged up some of my old sketchbooks when cleaning up, and decided to let AI have a crack at them. I was curious if they would improve the art, or change the concept entirely.

While these platforms can create images from simple text prompts, you can also input existing images as a “base.” I scanned my old drawings and fed them into the machine, and here are some of my results…

1. Rock Man with monster

I remember thinking this was a modern masterpiece. That shading! Those details! I don’t think the stone figure is anatomically correct — but in my defense, he’s supposed to be made from stone.

While he was meant to be fighting the monster beside him, I think DALL·E decided it was his pet python. I’m not sure what is happening at the right of the screen, though.

2. ‘Be Cool’

This early 90’s classic drew from my brief infatuation with smoking. I still think cigarettes look cool in drawings and art, but luckily I didn’t pick up the habit. Lung damage is not cool.

I included two generated variations for this one, from Midjourney and Stability AI (DreamStudio). One looks like a character out of Pulp Fiction, which I suppose is a fair interpretation of “be cool.” The other is a messy scribble, but I kind of like it.

I’d like to note that DALL·E flagged my drawing as potentially sensitive content. I assume it’s because of the cigarette. Scandalous!

3. Angel

I don’t really have a cool name for this piece. Anyways, you can see that I had to add a black bar across the breasts, as these AI image platforms are quite prudish in my experience, and flag anything that resembles a nipple.

Anyways, here are two generated versions of this one. The DALL·E version looks like a 3-minute life drawing sketch, but I do really like Midjourney’s take.

4. Scary clown reflection

Who’s not afraid of creepy clowns? Creepy people, that’s who. I drew this one sometime in the great 1990’s — clearly I was in a good mental state at the time.

I do like the way DALL·E followed the original quite closely, sans the eraser marks. It looks as if DreamStudio went with a more abstract approach — almost like how I see myself in a mirror.

5. Self portrait

In the 1990’s, we did not have social media or AI generators. Instead, we set up the self-timer for a photo of ourselves (on film), developed it, scanned it, and then drew from it. The original photo was of me looking like Bud Bundy sitting at my parent’s kitchen table. I added a silhouette in the background of the woman, who is actually waiting for me to move so she can order a drink.

DALL·E went with a sketchbook look, but not as detailed as my original. I’m not sure what Midjourney was thinking, but I love it. It’s sort of a selfie within a selfie.

6. Woman asleep beside ghost

If I recall correctly, this was supposed to be a woman mourning her dead lover, who was snuggling in beside her in spirit form. The only result that made any sense was from DALL·E, which put the (noticeably studlier) ghost on the opposite side of her.

The ghost has pulled down the sheets more than my original — naughty spirit! And check out her bed head. She’s not going to be happy when she wakes up.

7. Cabin by the river

Most of the drawings I came across were b/w pencil sketches. However, I found this watercolour scene complete with river and mountains, and decided to give it a whirl.

The best result for this one was from DreamStudio, which rendered the grass and water details quite nicely. I kind of wished it would’ve finished colouring in the background, as I was apparently too lazy to do at the time.

8. Elf in battle

Here’s another one I found half-coloured. Anyways, to my surprise, DALL·E went the extra mile and finished the colouring. Unfortunately, it changed from an elf with a sword to what looks like a human warrior with a spear through his head. ‘A’ for effort though, DALL·E!

I tried not to influence the results by adding specifics to the text prompts alongside the original images. To be honest, I’m still honing my skill at using images as a base, but it’s a fun experiment.

So, what do you think? Am I a better artist than these AI generators? Let me know!