I found myself deep into the Midjourney rabbit hole. And my goodness, how much I enjoy it :)
There's been a proliferation of creative AI lately. Dall-E, Midjourney, Stable Diffusion and Craiyon are maybe the first iterations of an entirely new creative ecosystem. And one that is already producing mindblowing results (and sparkling controversies).
For my latest video, I picked Midjourney because it seemed a perfect match to my style at first glance, but it doesn't mean I think Midjourney is better than the competitors.
By the way, you can follow me here (but you can only see what I made public)
Here, I showcase what Midjourney can do but also try to recreate one of my recent images (from the Fujifilm 150-600mm video) to see if this Artificial Intelligence can do a better job.
I will also show you some of my other creations, particularly landscape, portrait, and street photography, where I think the technology really shines.
But if the results are so good, is this the end of Photography? Hmmm… Not so fast!
Maybe it's because I'm an artist myself, but I see more potential than threat here. More in the video.
The Video
As usual, here's the direct link in case the embed doesn't work: https://www.youtube.com/watch?v=qAsVgiUHEFU
As you can see in the video, the AI works by digesting a few words (“prompts”) and returning what it “thinks” is the appropriate image. For all the images I share, I also include the prompts that were used, which seem to be very useful to everyone who wants to give this a go.
You should look at the Midjourney unofficial manual for the extra parameters, like aspect ratio, style, etc. They all make a difference in the end result!
I had great fun experimenting with Midjourney and making this video. Please enjoy!
Don't forget to subscribe. I'm definitely going to talk about AI often…
Updates
Dec 2022
Midjourney keeps getting updated much faster than the competitors. It's becoming the primary reference for AI Art, but it's also much harder to do anything with the platform without paying. I'm now on the highest tier because I'm paying the monthly subscription and the extra to have my art private rather than public.
Stable Diffusion would be the obvious choice to experiment with AI Art for free. But I haven't yet managed to achieve the same results. And if you're unfamiliar with some programming, installing and running all you need could be daunting. There's a nice app for Mac called DiffusionBee, but it's not updated to the latest versions of Stable Diffusion.
Anyway, going back to what I was trying to achieve in the video with the moonrise photo… Here is what Version 4 of Midjourney is now capable of, given the same exact prompts:
These are much better renditions, for sure. I particularly like the top right one.
The AI can now embed parts of the London skyline: some of the skyscrapers clearly resemble existing buildings. So it is definitely getting better, and I find it really exciting.
Dec 2023
With Midjourney 6, the improvements are even more noticeable. You can now clearly identify the skyline as London's. It's still clearly not real, but oh boy, it's getting close.
The AI is becoming smarter, but at the rate people upload photos online, there is obviously an ever-increasing pool of images to be used to train the AI.
Even here in my store, I offer my royalty-free photos for AI training purposes as well (for an additional fee).
I am now integrating AI creations into this blog. Whenever I feel the need to use stock images to complement my posts, I instead resort to AI. It takes me less time than searching for an image; it is more personal and obviously unique; and could be cheaper as well.
You have maybe seen these around already. Otherwise, you can keep exploring this blog… :)
See you next time! Cheers.
