Photobash in Midjourney without Photoshop! (Tutorial & Tips!)

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hey everyone today we are taking a deep dive into a mid-journey technique that I've talked about a few times on the channel here but I think it's really worth delving into because it's a lot of fun and it's fairly underutilized today we're looking at photobashing what it is how to do it and some Advanced ideas utilizing mid-journey's newer features and I should note that this is kind of a different technique than most of the other mid-journey photobashing tutorials that I've seen out there in that those tend to focus on generating images in mid-journey and then photobashing or copying them together in Photoshop afterwards whereas we're gonna kind of stick to Mid Journey for this okay let's get started so just by way of definition photobashing is a fairly newer art technique utilized heavily by concept artists or artists on a deadline it involves taking various photographs or images merging them together using editing software and then combining them to create one single cohesive image you tend to see a lot of photobash pieces in the early stages of development for movies and videos games on the mid-journey side we can take a fairly basic version of that concept and use that as a tool to collaborate with mid-journey for our final images and on the plus side you don't need to be a Photoshop expert or really even have a copy of Photoshop to be honest now that said you will need an image editing program of some kind if you don't have a copy of Photoshop there are alternatives both free and paid on the paid side I would recommend Affinity photo 2. I've used it in the past it's really great it's essentially Photoshop without the recurring subscription and on the free side there is a program called it's a long-running open source project it's been around for ages it's very stable and has everything that you could possibly need and finally on the free online side we have a site called Pixlr I have not played with it that much but it does appear to have everything that you would possibly need mostly we're looking for an image editor that has this layers ability as a note if you do decide to try out Pixlr just make sure that you pick this Pixlr e Advanced photo editor option the basic version is not going to have layers so at its core what we're doing is putting together rough collages for mid-journey to look at instead of utilizing multiple image prompts which can sometimes confuse the bot and you know we end up with unstable images and we're going to take a look at the bigger picture in just a minute but what I did want to showcase is how we can use photobashing for simpler and more subtle edits using very basic cut and paste techniques we can manipulate our images for really subtle edits it works especially well I think for acting and emoting something that mid-journey does not always nail for this example I generated a young woman having coffee at a coffee shop ar23 very simple overall it's a pretty decent image there are a couple of problems like the hand holding the cup of coffee like who holds a cup of coffee that way and she does have two different colored eyes so it's called tetrachromy I think tetrachromia um I don't know if that was intentional mid Journeys part or if that was a render during mistake we certainly did not call it out in the prompt but because we also did not call out any expression in the prompt she's kind of got that neutral you know mid-journey stare so I thought that this would make a good example to try to photobash a happy face on and see what we got so I went over to unsplash and I grabbed this image of a happy woman smiling and all you have to really do here is just grab her face just literally do the hackiest job of all time paste that image on top of our mid Journey output and yes I know it looks terrible and then just export the whole thing as a PNG back to Mid Journey all we have to do is upload the file hit enter and now we have a URL by right clicking and hitting copy link we can then run our original prompt with the URL of our reference image in front of it and as a result we get these images in which we can see her expression clearly has changed while still retaining the elements of the original photo like the same color of the sweater the same atmosphere of the cafe if there is one thing that I think I would like to experiment more with is maybe placement on the bashed face and raising that up a little I think that might have caused sort of that closeness of the coffee cup to the face particularly in that second image where it's essentially masking her smile but that's something to explore some other time one thing that I think that photobashing is very useful for in mid-journey is adding props to characters so for example example this is a random picture that I found I think that's Channing Tatum dressed as a cowboy and we're gonna grab him and put him into a jungle scene so to do that in Photoshop you can just use the quick selection tool which just allows you to do that I believe that affinity and should have something along those lines in a worst case scenario Adobe has a free background remover you just upload your image and then AI knocks out the background for you and then you can download your image from there so that's an option if you don't feel like futson with quick selection tools from there I grabbed a random stock photo of a jungle rainforest scene and copy pasted Cowboy Channing Tatum into it unfortunately in a save as mishap I ended up losing the original source file but as you can see I then grabbed a blue umbrella I think this is actually a beach umbrella like it's an outdoor umbrella not a person umbrella and then pasted that onto our Cowboy chain Tatum's hand badly as you might see as well and then added a volcano in the background because why not taking that entire image and then exporting it out as a PNG and then running it with the prompt cinematic still medium shot a cowboy in a jungle holding an umbrella in the background a volcano is erupting gives us this image which is obviously leagues better than our terribly photo bashed image now you might say that that is not Channing Tatum nor does it look like Channing Tatum and that is true mid-journey doesn't really do exact likenesses instead what it's taking a look at is the overall Arc Type of the character in which case Channing Tatum was dressed as a cowboy so we indeed have a cowboy it is interesting that mid-journey lost the volcano in this output I think it's kind of represented by these like God Rays coming down though the background has that kind of like smoky-ish vibe that kind of gives the the vibe that the volcano had erupted and now there's just sort of Ash flowing everywhere that said if we actually specifically want a volcano back there um I've got some ideas on that that we'll talk about later in the video since we were discussing archetypes earlier here's an interesting way of creating characters Within mid-journey uh that have archetypes that obviously mid-journey will recognize because it created it simply use the prompt imagine full body shot comma a description of the character comma Studio background comma white wall and then set it to an aspect ratio of 2 3 that way you ensure that you're getting a full body shot from head to toe and you'll get a character that mid Journey will recognize because you know it made it for example in this one we have a mystical warlock holding a sword uh in this one we have a viking yelling in battle swinging a war ax as a quick tip by the way someone in the comments had mentioned that they were having problems generating swords and axes generally if you put a action in front of that weapon uh that the character is doing like in the Vikings case swinging a war ax um you tend to have a better chance of getting it then there's also a businesswoman walking on a phone walking on a phone I put walking on a phone apparently not talking on a phone she's walking on the phone uh and then finally a soccer player bouncing his ball with uh his foot um though it does look like he's going to miss that shot as a follow-up on that aspect ratio of 2-3 if you do have a reference image that you want to use that is not in Q3 for example the uh Black Widow here in the superhero pose that's more in a landscape 16 9 format if you do choose to use that as an image reference and run your prompt my recommendation would be to not use an aspect ratio as I've noticed that you get a lot of cropped images that way like the head is missing or you know not missing but cut off rather or the feet are not in view so if you run it without an aspect ratio you'll get your Square image but you tend to have the full body in it that you can then cut out before we move on to the next section I do invite you to like And subscribe if you have not already additionally just as with the last couple of videos there is a PDF cheat sheet of the material covered in this video available over at gumroad it is free though if you would like to leave a donation you are more than welcome and I would like to thank everyone that has left a donation over the last couple of weeks it is very very much appreciated um yeah thank you very much okay let's get back to Bastion so once we have our character generated by May journey and again since we did it against that white background it should be fairly easy to mask out or knock that white out we can then create a background image for that character in this case I ended up going with a mystical Urban cyberpunk alley because I love mystical Urban cyberpunk alleys uh it did come out a little bit on the painterly side um but I was aiming for more of a cinematic thing that's okay because once we comp them together I can then run a prompt and then change it back into more of a cinematic style so I've put our warlock into our background image I didn't do the greatest job on masking um here in fact actually you can see this white background in between here and here I probably should have just used adobe's AI tool it probably would have been a lot more efficient but now that we have them together I can run a prompt and sort of merge everything together and give it a unified Style and ultimately ended up with this which I was pretty happy with you can see that in the background any of that Concepts sort of painterly look has been replaced with more of a photorealistic look because I called out a cinematic still in the prompt now that said it did take a number of Fairly terrible iterations before we got there there was this there were these images which maybe not so so terrible but I I didn't want that kind of painterly look and be it really feels like the characters are not actually in the environments you know it kind of has that it copy paste look some more examples where it's not quite working so I did want to illustrate that it's important to keep iterating on your idea and experimenting with your prompt because very rarely especially with image prompting and photobashing do things come out perfectly the first time out it's awesome when it does but it's pretty rare so that's just something that I wanted to mention since a lot of times in these YouTube type tutorials you tend to see only the best results and I just wanted to let you guys know you're not doing anything wrong we all have to go through many many many bad images to get to you know the one that we're looking for so anyhow moving on um so ultimately The Prompt that I ended up using for this image was the URL of our image cinematic still waited at four filmed by Ridley Scott Wesley Snipes as a mystical warlock comma wearing armor comma holding a wand while swinging a sword and then a multi-prompt break for mystical Urban cyberpunk alley and then an aspect ratio of 16 9. now that said our character did not come out very Wesley snipes-ish so let's return to the idea of photobashing and see what we get so nabbing an image of the day Walker himself blade um I just grabbed the top of him because I didn't need the rest of the trench coat and whatnot and then pasted it onto our wandering warlock pacing him in and again I did not have to do a great job I did at least try to line up the shoulders but again the angle is completely wrong that said I'm confident that mid Journey will suss it all out using our blade photo bash and running the same prompt got us this which I think is super cool um yeah maybe a little on the cheesy 90 side I guess but to me this looks awesome we could continue taking this image and photo bashing more and more but to be honest I was really happy with this I just think it's really cool another interesting thing you can do with your photobash images is run a describe on it so um again we took our Channing Tatum in a jungle with a blue umbrella image and let mid-journey figure out what all of that meant and the results that it came back with were pretty interesting one was a man with an umbrella near a tree in the style of cowboy imagery jungle Punk Ranger core data is photo montage another that I liked was man wearing a cowboy hat with a blue umbrella in the style of mysterious jungle and another one that called out uh steampunk and PS1 Graphics which was kind of a weird Choice mid-journey doesn't look very ps1y to me so while I don't think that describe is a good way of rendering your photobash I do think that it's an interesting way of exploring additional ideas to add to your prompt for your photo bash so I hope some of that was helpful to you if you have any questions please let me know in the comments below and if you have any photo bashing tips or tricks please let us all know in the comments below I thank you very much for watching my name Tim we'll see you soon foreign
Info
Channel: Theoretically Media
Views: 20,663
Rating: undefined out of 5
Keywords: midjourney, midjourney ai tutorial, midjourney ai, midjourney v5, digital art, ai art, photobash, midjourney photo editing, midjourney photorealistic, midjourney photo, midjourney photobash
Id: qeRVBKi1QMQ
Channel Id: undefined
Length: 13min 38sec (818 seconds)
Published: Wed Apr 19 2023
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.