Is Ai Image Generation the End or the Beginning for Graphic Artists?

[nelsonic]

I needed an intro image for some documentation I was writing for work. 🖼️ 📝
Searched all the standard "stock" image banks/sites and nothing good was coming up ... 🔍 ⏳
So I reached for the tool the "kids" are using ... 🤖 💡

Note: Not sharing the images generated for the work piece cause it's private (NDA, yada, yada ...)

Here's a sample image that can serve as the basis for discussion:
https://flux-ai.io/flux-ai/1830735307363778562/

Prompt:

Create a breathtaking, intricately detailed image of a modern architectural masterpiece,
showcasing the essence of [CONSTRUCTION].

- Style: Contemporary, photorealistic
- Setting: Urban landscape, surrounded by lush greenery and bustling city life
- Lighting: Golden hour, soft and warm, casting gentle shadows
- Mood: Dynamic and inspiring, evoking a sense of innovation and progress
- Color scheme: Sleek neutrals with accents of vibrant colors,
  enhancing the structure's unique features
- Composition: Emphasize the building's striking geometry and innovative design,
  capturing its interaction with the environment and the sky's reflection on its glass surfaces

High-quality 8K render, ultra-sharp focus, rich textures.
Ensure the image conveys the building's grandeur and modernity,
highlighting its role as a centerpiece in the urban fabric.
No text or watermarks.

This image is incredible. It's clearly influenced by many "training" images, i.e. the prior art created by humans.
But the question it raises is one worth thinking about:

Do AI Generated Images Spell "The End" for Graphic Artists?

Short answer: No. It's the beginning.
Longer answer: Maybe some people will be put off investing the time (years!) to develop the skills to create these sorts of images. But others will be inspired by the images that can be generated and motivated to learn.

The thing that struck me when using the tool is that it wasn't possible to edit the images that were generated.
i.e. if you want to make a minor change/tweak to something you can't!
Yes, you can tweak the prompt (input text) and will see a different outcome but not make subtle changes.
For that, for now, a human with graphic editing skills is required.
And since the images generated by "Ai" are renders, there is no source asset that can be edited.
In other words, these images are not meant to be edited, they are "final" renders.

Is it only a matter of time before we can highlight/circle an area of a generated image and prompt: "change this to that"?
Probably ... 💭 ⏳
And is the next logical step to feed the low-resolution renders into larger models and request higher resolution output? Yes.

Personal Context: I Love(d) Drawing ...

I used to love drawing and would spend hours on it in a flow state. ✍️ 🎨 👨‍🎨 ❤️
But one day (more than a decade ago now) I spent 3 hours on a drawing ...⏳
and when I realised how much time I had spend I decided to stop. 🛑
Not because I don't enjoy it, but because it wasn't "productive" ... 💭
i.e. I could be spending that time coding or building something that generates income. 💰
I reasoned: "I'm never going to get paid to be a professional artist, so why invest my time in it?" 🤷‍♂️
Long gone was the child-like pleasure of just enjoying the task and being fully immersed in just creating ... 😍
replace by the adult concerns of "how will I pay this month's rent?" 💸

I wonder what young people will do with this technology ... 💭
It's already being used maliciously/criminally ... https://www.bbc.com/search?q=deepfake
which, as a parent, is profoundly concerning. 😖

But the potential for good is still there. 💭

[nelsonic]

https://flux-ai.io/flux-ai/1830644601945260033/

https://flux-ai.io/flux-ai/1828100332151025666/

https://flux-ai.io/flux-ai/1826560194602250242/

Try it: https://flux-ai.io/flux-ai-image-generator/

[nelsonic]

Using this example to gen an image: https://flux-ai.io/flux-ai/1828915688986574849/

Style: Vibrant anime art, slightly stylized
Character: Stephen King with a thoughtful expression and twinkling eyes
Outfit: Cozy sweater with subtle books and quills pattern, hinting at his prolific writing 
Pose: Holding a pen and an old, mystical book, exuding creativity and mystery
Background: Subtle Maine landscape elements, like forests and Victorian houses
Symbols: Include small ink pots and floating ghostly spirits
Colors: Use rich, deep colors like navy blue, maroon, and forest green
Atmosphere: Convey an intriguing mix of spooky and creative energy
Ensure the image is visually appealing and captures Stephen King's essence in a fun, anime-inspired style.

[nelsonic]

This one I'm using for "work work":

Create a sticker.
Style: Vibrant anime art, cyberpunk stylized
Character: Greek Goddess Athena with a thoughtful expression, 
Outfit: smartly dressed wearing glasses, hinting prolific writing 
Pose: elegantly typing on a macbook air laptop, exuding creativity confidence, authority and reliability
Background: Subtle skyscraper city scape, sunrise
Symbols: currency symbols, banking, finance, graph
Colors: Use rich, deep colors like navy blue, gold, and forest green
Atmosphere: Convey an intriguing mix of creative energy and intelligence
Ensure the image is visually appealing and captures Greek Goddess essence in a fun, anime-inspired style.

Low-key this is incredible! 🤯

This is using the base flux.1 schnell model which is available on HuggingFace:
https://huggingface.co/black-forest-labs/FLUX.1-schnell

How long before Google image search is just AI-generated images and we lose a sense of "reality"? 💭 😬

[nelsonic]

https://medium.com/@researchgraph/the-ultimate-flux-1-hands-on-guide-067fc053fedd

[nelsonic]

https://www.reddit.com/r/StableDiffusion/comments/1f2mmcd/how_can_i_use_flux_locally/

[nelsonic]

https://deepinfra.com/blog/flux1-dev-guide