A Deep Dive into Midjourney & Historic Photography

hi everybody today we're going to take a look at one of the more controversial aspects of mid-journey namely realism and the bending of documentation through photography to me this is an area that has a lot of creative potential and unlocks the door to some really interesting creative techniques so I do think it's worth exploring so we'll be looking at how to use mid journey to recreate some of the most famous photographs in history as well as creating photos from historical events from a different angle and again most importantly we're going to be looking at some techniques to make everything look a little more authentic so the gap between mid-journing reality has been blurring more and more every day I think the most recent example being the Pope in the puffy jacket image that went viral a little while back and on the opposite side of the religious Spectrum there were some images of kids at Satan Con in Boston that caused a bit of an uproar enough so that the Associated Press actually had to issue a statement about it I mean to me both images are hilarious but I will also admit to having a pretty warped sense of hum humor more interestingly though was a project that I ran across by Michael Christopher Brown who is a legit award-winning National Geographic photographer there's no question that Michael has photographed some stunning images but his latest project 90 miles which documents the 2022 Mass Exodus from Cuba has caused quite a bit of a stir and you can probably guess why on the origins of the project Michael said I explored documenting the story in a variety of ways while on the ground though due to the secrecy and Trust necessary to the planning and departure from Cuba found access to be impossible as any documentation may have jeopardized the safety of those planning undertaking or in any way associated with the crossing considering himself a visual Storyteller Michael was undeterred by the method of image capture and instead utilized mid-journey if a photograph of Cubans arriving in Florida on a raft is on the front page of the New York Times this morning it will of course generate awareness but may likely be forgotten by the next morning as as we've seen these pictures for decades but when we create stories in new ways and are of course clear with our methodology it may help people see the story in a fresh way enough that it creates further conversation now to be clear Michael's images were never intended for news publication but rather as a storytelling device to bring awareness to the plight of these people but obviously it caused a bit of a Ruckus especially considering his pedigree as a photojournalist so all that got me thinking you know ethics aside how does mid-journey fair as a documentarian can it recreate or even recontextualize historical images and how would it fare against some of the Masters of Photography well I think with a little elbow grease it does really really well um let's take a look at that starting with a lunch break I think most of you know this photo titled Lunch Atop A Skyscraper it was taken on September 20th 1932 by an unknown photographer during the construction of the building that is now 30 Rockefeller Center unfortunately the identity of most of these men has been lost at time we know the names of about three of them including this lunatic at the end that is holding a bottle and I can guarantee you that is not a bottle of water he was eventually identified as Gustav popovic as after his passing this photo was found in his house with the following written on the back don't you worry my dear Mariska as you can see I am still with bottle cheers to you Gustav uh now let's go take a look at recreating this image in mid-journey and see how it does I started by running a describe on the image forward slash describe and ended up with the four generated prompts the images were okay I mean it got the idea of it but not the composition in some cases things got a little bit surreal which actually I really kind of like these I think they're really cool again not what we were looking for in terms of recreating a photo but a cool image nonetheless there's another set where we went completely to the other side of the world I kind of appreciate that mostly I would consider the describe outputs to be a strikeout so I decided to take a new approach by taking the actual photograph running it as an image prompt taking some elements from the described descriptions and then just being a little more straightforward in my prompt in the front end where I just said construction workers on the ledge of a building and ran it and we got these images of which I thought the third one was probably the most workable so I upscaled that and then brought it into Photoshop and just as a quick FYI if you want to follow along but you don't have a copy of Photoshop there are alternatives available I went over them in the mid-journey photobash video that I did that link will be below uh very briefly you can check out Affinity photo which is about a hundred dollars and it's a one and done thing there's no subscription to it whatsoever on the free side there is a software you can download called it's open source photo and imaging editing and manipulation it's been around forever it's really great and then lastly if you just want to handle everything online there is a website called Pixlr the links to everything will be down below one of the things that I noticed that was kind of off about this is that the colors are wrong despite it being a black and white photo this does have more of kind of a brownish sepia tone so one of the things you can do in Photoshop is just come up to image adjustments and then over to match color and then you can actually match the overall color spectrum of your original image to your mid-journey image so um just within a click of a button now to be completely honest I did spend a couple of more minutes playing around with the exposure and the Black Level the white level and a couple of other things but ultimately I did think that we did get more or less in the ballpark of the same overall color palette of the original there's obviously a lot more problems like the I mean the textures of the buildings on in real life are much better than the ones here and ultimately the characters do look kind of bland my trick for this is just to actually take a texture like a old photo texture and just pop it on top of it and just kind of messy the whole thing up a little bit which I think helps sell the image a bit by doing so you just grab a um if you Google photo texture you'll find something that looks like this and then you can just set your opacity mode to something like collar burn or multiply and just kind of play around with the various models and you'll find something that you like the ultimate problem with this photo is of course the characters I mean they are just sort of sad Bland blob AI people whereas in the original photo um the I mean these guys are characters just I mean if if you just spend time looking at their faces they're just they're such characters um especially Gustav look at him look at Gustav amazing there actually is a workaround for these sad blob people here and we're going to take a look at that in just a minute but first we're going to move over to another famous photograph and take a look at a very surprising mid-journey output there this is one of the most famous photos in American history taken by Dorothea Lang in 1936 this is migrant mother which has become an icon to represent the Great Depression Lang found the mother and her children pretty much by accident she was driving home on a cold and rainy March day in 1936 and passed by a sign that said P Pickers Camp She kept driving but couldn't get the thought of the sign out of her head so she ended up turning around and driving back 20 miles one of the most interesting aspects to me is the fact that this is actually one of six images and I think this sort of relates in a weird kind of way to mid-journey bear with me here Lang was a documentary photographer and preferred to Capture Moments in time without any sort of intervention or Direction Lang's first two images are kind of wider angle and from a distance and she realized that it lacked any kind of central Focus so she began to hone in on the mother and her child in the third and fourth image which I actually can't show you in full because she's breastfeeding in it and I mean we live in a stupid world what can I tell you you can go to Wikipedia and see the images uncensored if you would like uh anyhow the mother is nursing her child Lang sensed a sense of Shame and embarrassment from the mother's gaze and realized that this was not the image to capture in Lang's fourth and fifth image she tried a different technique in having one of the children come around to the mother's side and rest her chin on the mother's shoulder but with two small changes Lang elevates the photo into the icon that it is today simply by moving to the camera to the other side of the mother and then directing both children to look away from the camera thus avoiding competing gazes to the viewer we now have Migrant Mother the Takeaway on the mid-journey side for me at least is that you're probably not going to get your best image off your first prompt you should always be kind of experimenting with your prom trying out different you know camera shots and angles and really honing in on the image that you're looking for just keep experimenting and trying and eventually you'll find something great so once again using describe we run Migrant Mother interestingly in uh option two Dorothea Lang actually comes up as a prompt so it definitely identifies Dorothea Lang as the photographer of this image the results were interesting this one looks a little on the Hollywood glossy side to me and this one while beautiful is not at all what we were looking for but what is interesting is when I ran the describe prompt that called out Dorothea Lang and ended up with these images which yes are super surreal the kids are actually looking forward and they're not kids they're just clones of the mother um super weird but it's very very very close to what the actual photograph is so once again taking the Migrant Mother photo and using it as an image prompt and cribbing some lines from the previous described Generations we end up with this which is actually really surprising because if you were to look at that you would pretty much say that's just the actual photograph but it isn't if you look at a side by side with the actual photograph there are very subtle differences like the eye line is not the same obviously there are more wrinkles on the mid-journey output the mid Journey output has an overall like HDR kind of look to it as well but I think that could be toned down in Photoshop but it was just surprising to me because that's the first time that I've ever seen something come out so accurate from an image prompt I mean it's practically the same image and I would bet that if you were walking down the street stop 10 people and just showed them the mid-journey output they would probably say yeah that's that's the famous photo isn't it I don't know it's pretty crazy so just out of curiosity that was in 5.1 raw so I took the same image prompt and same prompt and ran it in 5.1 regular the more creative uh version of 5.1 and these were the images that we got which are still extremely close this gets in line I think closer to the idea of using mid-journey as an engine to create alternate angles to images that exist there are still some problems here like the you know this kid lives a clone of mom but this image I think with a little more work and taking some of the ideas that we used in Lunch Atop A Skyscraper you might be able to fool people into thinking that this was like a seventh photo that Dorothea Lang had that day in 1938 this is Ansel Adams Monolith the half face of Dome one of his most famous photographs and was shot at Yosemite National Park in 1927. it is stunning it's unfortunate that ansel's work has become so synonymous with doctors waiting rooms or sort of benign desktop wallpaper that it's easy to forget what a masterfully composed technically brilliant and just downright patient photo this really is Adam's hiked for half a day for this Photograph carrying 40 pounds of camera gear on his back and I think the elevation here is like 3500 feet so yeah he hiked up carrying all of that stuff and even when he got there he spent time experimenting with different filters and ultimately ended up waiting two and a half hours for the sun to be in the right position for this photo I typed some words into my journey what are you gonna do so describe actually didn't call out Ansel Adams which was surprising to me considering how iconic this photo is um the results that it did return to us were interesting this was very nice it was very reminiscent of the Mac OS Yosemite default wallpaper we did end up with this bizarre one which kind of looks like the set of an 80s sci-fi B movie some other interesting ones kind of a bridge to nowhere sort of thing ultimately though I don't think that you ever really attain what you're looking for unless you use your initial image as an image prompt in this case we finally did end up with something that more or less resembled Ansel Adams work but even as it succeeds it also kind of fails like I think it did a really good job of capturing the overall tone of the sky that's actually a technique that Ansel used by putting a red filter over the camera lens when shooting black and white it kind of blocks the red light spectrum and causes that really nice deep saturated color but it fails on the textures of the mouth Mountain itself to me it doesn't track as realistic it kind of has a video game like Unreal Engine kind of quality to it even though I think actually Unreal Engine has gotten even better than this so I can't necessarily even say that but I do have a potential fix for this so once again just by taking some sections of the original photograph copying them and then pasting them on in this case I used the soft light mode here um and then this side was kind of problematic so I dropped another one there once again soft light both had a 100 it kind of now buys just a little bit more so with just some very basic edits we now have something that I think you could print out hang up in a doctor's office and sadly people would just continue to ignore going back to the 90 miles project I did get to thinking what about photographs of historical events from different Vantage points weirdly enough I started with JFK so I was at a buddy's house a couple nights ago and for some reason Oliver Stone's JFK was on TV and I caught some bits and pieces of it and so I think that was sort of in my head I also think that when I started thinking historical events from different Vantage points immediately the grassy knoll popped into my head but I didn't want to go in that direction so instead I just went with the presidential motorcade which gave me some difficulty a lot of the initial images either had Kennedy driving the car itself in this image there was a Ghost Driver it would often forget Jackie Kennedy and in this case apparently they are also in the UK as the drivers on the wrong side or in this image where it's JFK sort of partying and there's a drunk frat boy in the back seat taking a page from Dorothea Lang I began to switch up the prompting and moved into a close-up which resulted in this image which I actually think is pretty good there are some problems where it kind of has a wax figure sort of look to it but let's see if we can't dial that out in Photoshop so in this case I just made it black and white and turned up the grain a bit and then ultimately ended up adding a level full of 35 millimeter film grain to it and then lastly applied another filter to it which led us to this image um which I think actually looks pretty on point I don't know you guys let me know does that actually look like a legitimate picture of John F Kennedy or can you still see AI within it keeping on with the theme of assassinated presidents so I'm looking forward to seeing the FBI at my house later tonight um we moved over to Lincoln I thought it'd be an interesting idea to see if we could get a photograph of Lincoln at the Gettysburg address so I always thought that there wasn't an image of Lincoln giving the Gettysburg address but in researching how I was going to tackle this image I found there is a photo of him giving the Gettysburg Address it's just it's this photo which is not very good so let's see if we can do a better job of capturing this moment in history using mid-journey as a side note I also learned that there are 150 photographs of Abraham Lincoln that's a number that sort of surprises me I don't know if it should be higher or lower it just sort of surprises me that there are 150 photos of Lincoln a lot of the images of Lincoln were shot on the daguero type style of camera in my initial prompt I tried out a photo of Abraham Lincoln Gettysburg Address daguerreotype quarter plate and ended up with these images which that's not quite what I was looking for moving over to a photo of Abraham Lincoln addressing a large crowd of Union Soldiers Gettysburg Address daguerreotype aged photo got us this which which looks okay but doesn't really pass for reality to me the crowd is far too large um there's a giant Amphitheater in that fourth image yeah it just doesn't buy for me so what was surprising to me was taking that original image of the Gettysburg Address out of focus blurry and poorly composed as it is and running it as an image prompt along with my prompt which resulted in this image which I actually feel kinda Works ultimately there was still the problem that if you zoom in he's kind of got that mid-journey distorted facial feature thing happening uh luckily for us of the 150 photos that Lincoln was in we actually do have a profile shot so taking this photo and just cutting out his face flopping it around and then applying it to our mid-journey image we get this which I actually think kind of works I probably should have spent a little more time matching the skin tones there but I think it more or less works and then if you beat the whole thing up a little bit and get to here I think you've got something that might fool people um I think there probably is a level of grain that I probably should throw over his face and maybe blur it a little bit more but I would say it's 90 there I don't know you guys let me know do you think this is possible do you think that if you were just to show this to a couple of random people they would say yep that's Lincoln giving the Gettysburg Address so what's this all mean can we no longer trust anything we see do we really live in a post-truth world well yes and no I mean honestly photo manipulation has been around for a while now and to be honest there's nothing that we did today that couldn't have been accomplished by Photoshop over the last 20 years and as far as the truth and reality of a photo well all photos are a manipulation of reality any event photographer can make the most boring party in the world look like the biggest Banger or a wedding look like a funeral all through shot composition shot selection and post-processing Dorothea Lang posed those kids in what is now considered to be one of the greatest photos of all time there's never truth in an image there never has been as a medium photography is less than 200 years old which actually makes it pretty young in terms of the Arts AI imagery is what two now I mean it's an infant it's not even an infant it's a newborn and like with all newborns we do have to be careful with it but if we raise it right it might just grow up to be an incredible artist hey thank you for watching I think this one ran pretty long so I really do appreciate you sticking around to the end my name is Tim we'll see you soon [Music]

Channel: Theoretically Media

Views: 6,233

Keywords: midjourney, midjourney ai tutorial, midjourney prompts, midjourney history, midjourney historical, DocumentaryPhotography, AIImagery, RealityVsAI, recreating photos, post photography, Ai Art, ai photography, ai photo editing, ai photo generator, midjourney ai, midjourney tutorial, midjourney 5.1

Id: OmdzBXxvDnw

Length: 19min 48sec (1188 seconds)

Published: Sun May 14 2023