Photobash in Midjourney without Photoshop! (Tutorial & Tips!)
ฝัง
- เผยแพร่เมื่อ 18 เม.ย. 2023
- Let's explore how to Photobash in Midjourney, without using Photoshop. In this video, I'll take you through the concept of Photobashing and show you how you can create your own images to use as image prompts with Midjourney. And we'll do it without needing to know Photoshop, or even needing to own photoshop!
Of course, you might be wondering, why should I bother with Photobashing when I can just text prompt? Well, I go through some of those reasons in the video, but you'll quickly see that the technique is very useful for everything from subtle edits to creating characters and scenes!
FREE PDF CHEATSHEET: theoreticallymedia.gumroad.co...
LINKS
Affinity Photo: affinity.serif.com/en-us/photo/
GIMP: www.gimp.org/
Adobe Background Remover: www.adobe.com/express/feature...
Follow Me on Twitter: / theomediaai
-------------------------------------------------
Thanks for watching Theoretically Media! I cover a large range of topics here on the Creative AI space, Technology, Tutorials, and Reviews! Please enjoy your time here, and subscribe!
Your comments mean a LOT to me, and I read and try to respond to every one of them, so please do drop any thoughts, suggestions, questions, or topic requests! - วิทยาศาสตร์และเทคโนโลยี
![จังหวะนี้ - เม้ก อภิสิทธิ์ [ Official MV ] จอนนี่มิวสิค](http://i.ytimg.com/vi/4gQRZCkRyiU/mqdefault.jpg)
It's great that you mentioned that some channels only show the best results immediately, but in fact we need to go through many iterations. Appreciate the transparency
Thanks so much for that! That's something that I try to get in for every video. I think it's super important to reenforce that idea, since far too often I see beginners getting discouraged with their initial results and blaming themselves. I also think it also leads to the perception that generating AI assets is an "easy button"-- and it can be, if you aren't looking for something specific-- but when you are, it (like everything else in life) takes work!
btw, i tested a quick /prefer option set for character gen opt using your prompt, and it worked great:
/prefer option set option as: name value as: full body shot, studio background, white wall --ar 2:3
In this case "name" is whatever you want the command to be. I used chargen meaning character generation because it is easy to remember.
Use it like this: /imagine girl wearing a pink dress --chargen
That's it. It will produce a grid of a girl wearing a pink dress in front of a white studio wall as a full body shot at ar 2:3
Now you can use --chargen as a short cut to produce any number of characters in this setting.
Pretty cool tip. Thanks for sharing the idea.
Oh, that’s a great idea! I’ll explore that soon! Fantastic tip!
Great insights. I just tried using "Photobash" as a style command. It came up with some impressive results that allow the combination of multiple elements to appear in the same image together without conflict. Midjourney should incorporate it as a workaround function to include a number of specific items, characters or elements in a scene. Currently very hard to achieve, otherwise.
That would be really awesome. I have heard that they are working on a "Pose To Image" type thing, where we can actually use the same pose as an image reference in our output. That alone would be a game changer!
That's working pretty awesome! Thanks a lot for posting. Even works if the final image is supposed to be a painting. Makes it look painterly, then without "photobashed"
Excellent content - I learn so much from every video. Thank you!
Thank you! That means so much, particularly as I'm sitting here on a weekend and planning out the new content for next week! (Honestly, don't feel bad-- it's super grey, cold, and raining outside, so I'd rather be doing this anyhow!)
As someone who has always been kind of intimidated by Photoshop, this is exactly what I have been looking for :)
It’s actually what I like much about it! You can take a very caveman approach to PS (or any photo editor) and get some stellar results!
Such a great tutorial! Very detailed. I really love the prompts that you displayed as overlay in large letters. What a great idea. I haven’t seen that in other tutorials thank you. Oh and a cheat sheet wow.
Fantastic to hear! And I’m glad the cheat sheet was helpful too! At some point, I’ll start adding the upscaled images along with the cheat sheet as well, that’s been another request. I just need to get organized a little better. After these videos, my desktop is a totally disaster of random icons, ha!
Thank you very much, though!!
Tim :) Terrific as always. Thanks again for the reminder about going through the process to get the images (i often get frustrated and forget, that the process I'm going through is normal and as you say "I'm not necessarily doing it wrong"), I'm looking forward to incorporating more photobashing, and adding this to my midjourney "tool kit", I think the addition of the describe feature will also add to the power and effectiveness of the technique and my results! Keep them coming, you're doing great work here!
Thanks so much! And yeah, I’ll call it out as often as I can. I’ve seen too many YT tutorials (on all manner of topics) which have the appearance of “do this, so that, voila Masterpiece!”
While I totally understand editing for brevity. It can also be a little disingenuous.
That said, really happy to hear you found the video useful!
the problem with "cyberpunk" is most of it's artwork in MJ training are digital paintings from places like artstation. You might get a more realistic looking result using a photographer like Liam Wong, who takes fantastic neon drenched photo of Tokyo. Also a little tip, you can, instead of right clicking on an image and selecting copy, just click and drag it to the message box.
Just checked out Liam Wong-- that is FANTASTIC work! I'll give that a shot next time around. Actually, on the "to-do" list is to create a series of cyberpunk images that take inspiration from different elements. As much as I love the aesthetic of "traditional" cyberpunk, I really appreciated what Villeneuve did with Blade Runner 2049, and (although mixed in terms of product) what CDPR brought to the genre with Cyberpunk 2077.
I love the fact that with MJ, we can explore and iterate on ideas like this!
Link to the PDF Cheatsheet: theoreticallymedia.gumroad.com/l/photobash
👋
Nice job, Tim... as always!
Thank you so much!! Still gotta get to that dash dash video! So much AI to cover!!
Amazing! I never thought about this technic to do so. Very helpful, thank you very much!
Really happy to hear!! You're welcome!
Even with access to photoshop and very extensive photoshop experience, I actually photobash in procreate 95% of the time- it usually loads and runs much faster and imo has the most user friendly interface, especially for beginners ☺️
Ah! I forgot about procreate! I should have mentioned that one. I know, Procreate seems to be very popular amongst artists who use tablets. And younger artists as well. I taught some digital classes w/ high school aged kids and they ALL were about Procreate.
Way cheaper too!
@@TheoreticallyMedia it sure is! I learned in adobe suite when it first came out and eventually got so used to the interface that the simplicity of procreate was intimidating at first….. now I totally get why younger people/those new to digital art start in procreate and stay there, truly less is more for many many applications!
@@E42545 Agreed! that's actually been an issue I've had with Adobe's subscription service. They now have to constantly incorporate new tools to justify the annual sub-- it just leads to a lot of bloat and useless tools!
This is awesome, Tim. i want to let you know you've done great education video!!
Thank you so much! Really appreciate that!!
I'm glad you made the photobashing vid. Another tool in the Midjourney toolbelt! Thanks
They're pushing so many cool things out, we're going to need to build a shed for all these tools!!
This is pure gold, going to try it
Let me know how it goes!
Awesome! Thanks for sharing
Excellent! Glad it was helpful, and thank you for the comment! It’s always great to hear!
Very informative. Thank you.
Excellent to hear! Thank you for the watch and the comment!
Love your videos man. I’m just learning midnourney and your content is great. Thank y ou.
Oh excellent! Thank you so much! You're probably past it now, but I do have a "Getting Started with Midjourney" video in the pipeline-- granted, with my schedule, we'll be on v6 by the time I get to it! haha.
@@TheoreticallyMedia lol!! I will most definitely watch it when it comes out, regardless of the version we’re on. Because of your videos I realized I could’ve been using v5 this whole time lol!!
Glad I found your channel, nice tutorial
Glad you found it as well! Thank you very much! And as a fellow Tim, I welcome you!
This one is new to me thank you 🙏 🙏
Excellent to hear! Hope you get some mileage out of it!
This is amazing!! 😮🤯
Thank you so much!! It was long overdue, I really wanted to dive in on the Photobashing technique. So glad you liked it!
Thank you million times for your work and videos!! You are amazing! Your videos helped me to understand the world of MJ, I already got an awesome project using AI, and thank you for the Adobe Background Remover, it is so refreshing to use smth that is not a Ps)))
That's amazing to hear! Thank you so much for your comment! And yeah, that background remover is clutch! It is a real time saver!
Super helpful! thx
Excellent to hear! Thank you!!
Excellent!
Thank you!! Glad you found it useful!!
Very cool!
Thank you! And thank you for the watch and the comment! Much appreciated!
That's a Starbucks umbrella Tatum is holding.
Good tie in to the first idea.
100% Tatum is totally addicted to the caramel frappe! It's amazing he stays in the shape he does!
@@TheoreticallyMedia hahah! I'm that way for caramel macchiatos!
MJ's Describe function is fascinating, a peak into its mentality. like many highly visual Talents, MJ is a bit off with words, in a charming way.
It really is. It’s taught me about some really cool artists by name dropping too. Half the time I end up googling stuff it describes, and I’m usually blown away!
Very cool
Thank you! Really happy you found it helpful!
Thanks!
You’re welcome! Thank you for watching!
THank you!
You’re very welcome!!
exactly what I need! Thanx!! I try to figure out how to better control exact perspectives in MJ. Perhaps that would be a nice YT video?
I’ve actually got that one on the docket! Maybe this week!
amazing tutorial, thanks Tim! Have you ever compared a super fast and rough photobash with one that is also quick but maybe a little better cutted out and placed in the composition? my thought was that you might need less rerolls to get the result you want... would be interesting if it makes a difference from your experience!
also have you experimented with this technique and -iw?
thx again and keep it up! great content
I’ll give it a shot! And yeah, I did a bunch of -iw experiments, but I kept them out of this video since I didn’t want to overload newer users. I’ll slide a mini lesson on it in an upcoming video!
@@TheoreticallyMedia thank you so much!
when you do this, be careful not make a soft reference image when trying to blend images using the eraser tool or gradient mask. It will result in a soft midjourney render. Same thing goes for using low res images in areas of your photobash.
good call! That sounds like some experience talking!
I don’t do prompts that require characters to look like anyone, as generally my characters need to be fairly generic, but I remember a few months back using the “lookalike” descriptor when first playing around with Midjourney, which seemed to do a decent job. i.e a lookalike of Wesley snipes as a cyberpunk warrior etc. Not sure if it still works in v5, and the versions of v4 still available.
I’ll try it out! I’ve also had success calling out “young (actors name)” it works particularly well with the 80s actors, since you actually end up getting them in their prime. Curious to try out lookalike!
Would love to see that prompt at 11:17 but with swinging a sword higher up the sentence. Im finding the order of a prompt can alter the outcome huge
wait? 11:17? I think that's "Wesley Snipes" on the first go around. Is that the one you were talking about? But totally agree: The prompt order is a pretty big deal. I also did a video where I took a look at how tokens work with prompting and the Midjourney bot here: th-cam.com/video/cH8UdeaYQls/w-d-xo.html
And I think that's another aspect of getting MJ to output what you're looking for!
Would you happen to know what the prompt should be if you want the character you're having Midjourney create give you a full-body pose sheet for 2d animation (front t-pose, 45 degree, and from the back)? Thanks in advance...
Try something like "Character Turnaround, 2d Animation, T-Pose, (details about character)" I'm not 100% sure the T-pose part will work, but I have done character turnarounds in the past, and that always works.
Terrific! I'm curious about using png. Does it work better than jpg?
I gotta say, for MJ, not really. Since I don’t think MJ really cares about the quality of the image, it’s just looking at the content of the image.
Mostly I go PNG because the images tend to be smaller sizes. Also, the format is handy since you can save transparent, whereas jpgs will bake a background in
haha, i did the face paste 4:00 onto a pic that had two people in it. i used the same seed on the new generation, and the resultant photo made both people have faces very similar to the face I pasted in. This happened in all 4 pics in the tile. i guess if you ever need a pic of twins...
Oh, that's happened a bunch to me too. Ha, that and if you try /blend you end up with some really messed up looking people!
Krita is a free image editor worth taking a look at as well.
Krita or Krea?
@@TheoreticallyMedia Krita. It's open source/free does 90+ percent of what Photoshop or CSP does. It might be worth a look...
Thank you! best source for Ai art education
Ahhh- bless you! Thank you so much!
I also tried the same workflow and got some decent results but seems like it only applies to a single-character image, when I create two or more characters and I want them to show up in one image Midjoourney always got it wrong, I tried many methods to solve it but it just didn't work out
The two character thing can be tough, when you’re looking for specifics. I think it’s a limitation of the amount of memory you get per image. I’ve usually found that you get one character who looks good, and the other looks like a wax figure!
Oddly, you can pull off crowds pretty well though!
Can't wait to see what v6 will do for this bashing.
Same!! They've been rolling out so many new features, I have the feeling that v6 is going to be pretty special.
@@TheoreticallyMedia my guess is that photoshop (in the sense of photobashing) could face a big dent if Midjourney would allow us to write "take the face of photo 1 and place it onto a person int a cyberpunk scene" .
@@davyvettorato3728 To be honest, I wouldn't be surprised if Adobe didn't introduce that into Firefly when it hits Photoshop.
I was using the photobash technique to get a friends face onto the body of a shark. Midjourney just could not understand what I was trying to do. Ended up with images of a mans face in front of a shark. Just couldnt find out of a way to come close. Would like to see you try something like that
Yeah, that can be a tough one. My overall thought is to go with a cartoon style over a photorealistic one. I think you might have the best shot there. Something like, Style by Pixar Movie- or maybe even call out Nemo?
I have a commission I've been absolutely struggling with, that I think this is going to help me with greatly.
I sure hope so. Even if it doesn't, this does seem very interesting to take a few fast hours and explore it..
-The Midjourneyman
If you’re running into trouble, feel free to stop by the discord and drop a line in the Project Help channel!
@@TheoreticallyMedia you bet! I love discussing things like this anyways, it'll be great to be around like minded folks.
Hahaha that Wesley Snipes is so f🌈bulous
He’s out to slay…vampires that is!!!
problem which I have not seen anyone overcome: put the 2nd woman's face onto first womans' and keep exactly same photo (AI changes entire image). Any ideas?
when midjourney will fix this hands problem?
V5 more or less has it down. You still get some wonk from time to time, but for the most part it’s pretty solid.
There was a pretty interesting video on Vox that detailed why artists, and AI struggle with hands. I can link you if you’re interested!
I cant see copy the image link, mine says copy image link!
Wait, when you right click on it, you can’t see Copy Image Link?
Dang! you found out my secret...lol
Sorry I let everyone know! LOL-- It really is a powerful tool!
@@TheoreticallyMedia I know i was just joking with you. Yes, it definitely is... wish I could share my latest work with you. I love criticism. I find that criticism makes you grow.
My results all look extremely bad photobashed 😢 It just copies the bad look too much 😅 I played around with stylisation and Image weight, the results looked better, but much to far away from my goal and original motif
Yeah, that’s kind of why I mentioned that part in the back half of the video, where I showcased the “bad” versions. My best advice is to keep experimenting. One idea is to take one of your outputs and then use that as a new image reference.
So, if you have one that looks “ok” but not totally up to snuff, upscale that and run it.
It’s moving one generation from your original bash, but it’s the end result that matters!
Photobasing with no direction what's the point.
To be honest, I look at it more like you are providing MJ with visual descriptions for your image. Almost like creating a custom reference image.
so sad and lame
Thou shalt not steal! Those that live with AI will die by AI. Thou shalt not steal.
Hi @TheoreticallyMedia, thank you for this video, could you say what is the first one free program that we can use because I was able to identify since my English is limited and in transcription it does not appear. Sorry, do not need, I saw in description is Gimp.
correct, it's called GIMP. You can download it here:
www.gimp.org/
Hi @TheoreticallyMedia, I repeat your experience, the Wesley Snipes in alley, believe I did with Bluewillow and works, I'm stoned. Thank you very much.
Great to hear! I’ve had a lot of mixed results with BlueWillow. And…did you say you were stoned? Haha. All good!