One could argue that generating portraits with Midjourney is actually harder than taking real photographs, since photographers spend only milliseconds taking their photos whereas prompt engineers sometimes spend hours copying text from PromptHero or Lexica and pasting it into Discord.
And building software is easy, you just have to type `make build`.
Like in development, what's hard in portraiting is not pressing the button; it's spending weeks to find the model, dress the model, groom the model, do the model make-up, find a good spot, find a good lighting, find a good angle, chose the best compromise in all the settings, and so on.
The prompt engineer has to explicitly set each parameter, he cant just press submit like the photographer presses the shoot button so I agree with you.
> One could argue that generating portraits with Midjourney is actually harder than taking real photographs, since photographers spend only milliseconds taking their photos whereas prompt engineers sometimes spend hours copying text from PromptHero or Lexica and pasting it into Discord.
LOL...
- a photographer needs to find the right model, it just can't 3D print it.
- the model needs the right clothes, you just can't 3D print that
- the model needs the right make up
- the model needs the right haircut.
- the photographer needs to make the model take the pose he envisioned.
- the photographer needs to chose the right spot to take a picture
- the photographer needs to take that picture at the right time of a day
- the photographer needs to chose the right lenses
- the photographer needs to get the right lighting equipment, place the lights correctly and so on and so forth.
and I'm sparing you the details of all the steps in post production...
in the meantime, a magical prompt did generate 10,000 images as your alleged hard working "prompt engineer" simply had to enter a few sentences in a text area or just copy paste another prompt with slight changes, zero photograph experience needed.
So no, being a photographer is much harder (and expensive) than using midjourney.
This process to me is not really photography. They're making a photorealistic composition. Why bother with all that work, getting the conditions right, when you're going to photoshop it afterwards?
Cue sunshine telling me how all photographic images have post-shot interpretative processing(?).
When I use tools to create a composite where the disk of the sun is visible through clouds, and the back of a mountain below is added from a second shot, is this not photography? Because it allows the viewer to see a scene a camera with its limited dynamic range cannot capture in one shot. Yes the editing may take hours.
Grayscale is not very realistic, is it? Yet we adapt our perception and agree that it captures some reality.
One can use the tools to create a representation of the scene so that observers would agree this is what they saw. And one can use the tools to create a new scene from pieces.
I like to take pictures of scenery with the moon in it. I have no problem using different contrast settings for the moon. But I wouldn't allow myself to move the moon around in the image. Why not?
When we use computers to create faces combining millions of shots, it can be argued that this is photography too. (Though I do struggle to use the term here.) The deception comes in only when we present the scene as having existed on its own.
I couldn't help myself while reading the other comments, so I asked ChatGPT the following prompt, in an attempt to fit in:
> write a reply to the following comment, in a style of a stuck up hacker news commenter. Ridicule the idea that photography is easier than creating AI art. Point out how much effort a photographer needs to put in: (your comment)
Result:
Oh dear, it seems like we have another individual who's completely missed the point. Let me break it down for you, dear commenter. Photography isn't simply about "spending milliseconds taking photos." It requires years of practice and honing one's skills to capture the perfect shot, understanding of lighting, composition, and an eye for detail that can only be developed over time.
Meanwhile, creating AI-generated art with Midjourney or any other software is just a matter of feeding the right data and parameters to the machine. It may require some technical knowledge, but it certainly doesn't require the same level of artistic skill and intuition that photographers need to possess.
So, if you really think that copying text from PromptHero or Lexica and pasting it into Discord is the equivalent of the amount of effort and artistic vision that photographers put into their work, then I'm afraid you've got another thing coming. Perhaps it's time to put down your keyboard and pick up a camera, so you can truly appreciate the art and effort that goes into photography.
---
I can't wait for the top story next week where one of us comes clean as not actually being an expert programmer and a successful entrepreneur!
It took me 3 tries to "engineer" a prompt that'd produce a sensible reply to this, so one could argue that it is indeed harder than, say, understanding sarcasm on the first try.
1. Use a python script to iterate over model,seeds and prompt (pos/neg) variations
2. Use the resulting image in img2img or ControlNet for further processing
3. Use photoshop for finalizations.
