Where did the thumbnail, title or intro promise you a lesson on character consistency? Exactly, it didn't. You're projecting your assumptions and blaming me for that. How does that make sense? Right, it doesn't. My "audience" knows that I have an entire playlist on character design and the end screen of this video even promotes it. You, the random viewer may not know this, but then you're technically not really my audience. So I suggest you direct your frustrations elsewhere or inward. You're getting a free education here, so I recommend dialing down the sense of entitlement. It clearly isn't doing anything for your mental health. This video has a like ratio far beyond 99%, so clearly the video isn't the problem. Yes, that's a hint for you...
Looking for consistent characters? Big fan of the channel, Christian. We've used your videos for training staff. I suspect a lot of people will get hung up on this even though you explained it a bunch of times...haha.... If you're new, go see the earlier videos on character building, we're doing this stuff professionally (I have a small staff doing character creations for video shorts) and this channel has been our go-to channel for midjourney stuff. That said, the "I want a quick tip" verison can be summed up like this: In the first step, don't use any background (I think Christian did this to make a point, but it's actually more work and can create other problems later). Creating this character on a plain background also has the added advantage that you can remove the background and use the character on a transparency layer to add to a scene (like the one he wanted 3 characters but knew that MJ couldn't do that correctly). Next, once you have a character or character you like, save the URLs for that character (whether it's 2 or 15 URLs, just save the short links). Then, instead of prompting like this: "An asian woman in her blah, blah, blah, blah" over and over, just prompt: A woman URL1 URL2 URL3 URL4 etc, (as many as you need) in SITUATION and SCENE ( in a boat on an Amazonian river or in an Asian jungle or fighting a snake or holding a map or entering a cave, whatever....) IN THE STYLE OF and then whatever you're using here to make the style you want (you can cut and paste this instruction over and over for a consistent style, whatever that is). And there you go... As a final note: You can make adjustments to the style and chaos while leaving the URLs and the SITUATION and SCENE and IN THE STYLE OF instructions identical and get much different results, so it's worth playing with those too (not just rerolling). One other thing I've noticed, MJ does seem to make changes in style and look when you use a different aspect ratio. So, for example, sometimes I'll use --ar 2:1 even though I want a square panel because the look is better for that scene and I just trim off the sides, so don't forget that you don't have to make the aspect ratio in MJ the aspect ratio you need for a panel or page or scene, you can make it different to get the look you want and then cut it out in PS or GIMP or whatever to the right size. Sometimes this is less work than trying to force MJ to create a certain image in a certain aspect ratio. Imho. Thanks again, Christian, you've been a monster-sized help to us.
Depends on what you mean by 100%. Of course, the more stylized it is, the more variations you'll get. And I'm not claiming every single iteration is consistent, but if you run the prompt a few times with the same URLs (the number of them depends on what you're doing) you'll get a consistent character 100% of the time you're willing to put in the work to produce a consistent character.
@@TanChoonHong I'm finding that the situation will change the character when the training model's images can't figure out what to do, so, for example, using this model to put a character on a horse doesn't seem to work very well.....it knows the character, and it stays consistent, and it knows the word "horse" and will even put a horse in the image, but "riding" a horse...I can't quite get there.. That said, I'm getting very consistent images by using a group of URLs that force the AI to stay in a very narrow range of possibilities and sometimes it will put them where I want them. A lot of what I'm doing requires me to make a consistent character that is happy or sad or sleepy or whatever on a plain/white/transparent background and then layer that over the scene I want. If you try to use the method I mentioned above, well, I sort of jumped the gun on what it can really do, which is why, I suspect Christian didn't even bother with it, it's too hard when the scenes are complex to get everything right, better to make multiple images, and combine them in PS or GIMP or whatever. Good luck!
@@TokenizedAI I actually tried to "just go learn to draw". But I just suck at it. Best way to get "'Exactly" what I want would be to draw it myself. Still sucks when using a tool that's 1000% better at drawing that still takes 30+ prompts to "Almost" get what I ask for. Feels like my brain is on fire when I'm constantly changing and modifying my prompts but getting anything but what I need. Plus I've found that 90% of the videos on this is just having A.I do 100% of the work. Is it really bad that I want to use A.I to create a comic book for "My Stories" because I am truly horrible at drawing. In short, I have tons of stories to tell. Just no talent in drawing or thousands of dollars to hire artist. I wanted to use gofundme, but would be impossible without something to show process. Can't even start without some sort of funding, but don't think i would get far on promises alone. Considering all the scams out there.
That's true. And I said it before and I will say it again. This was only meant as a demo of how these tools can be leveraged. It wasn't meant to be a final product.
The consistency in the characters images are maybe possible with Insight FaceSwap ... creating first the stories images with Midjourney and then swapping faces .... I tested it rudimentery and it looks like a possibility. If you are good in storyboard and want to stand out of the drawing crowd it could be a possibility. Anyhow ... I think in a year we will talk advanced situations with this ... 🙂
Christian, you have given me a breath of fresh ideas and concepts that will enhance my journey in making good stories and images. I have been looking for an educator such as yourself, and I thank you!
one thing that you could do is draw the pictures the way you want them in a bad/rough way but with all the right composition and upload them as references for midjourney to create around. would probably turn out really good
Thank you for the lead. I've been trying to make consistent characters in v5 and the only thing that has got me 90-95% close is pasting the url of the target image and then following it with the descriptive prompt, only you have to maintain the identity names of the character/s you are generating.
I found your video interesting. The visuals look very impressive and look like comic art. I find that fascinating. However, I draw comics and I create storyboards professionally, there's more to comics than just "characters standing/sitting/looking" in an environment. There's acting, and pacing. There's....well...visual storytelling that needs to happen. A good test is if you can look through the visuals of a comic without the dialogue or the narration and still get the 80% of the story. Where were the puzzle solving shots? The moments of friendship? The visual collaboration? The emotional beats? The action moments? The danger moments? The resolution moments. All I saw where images of people in midshots mostly looking at the middle distance. I also didn't see an example of the final with dialogue balloons. My point is, yes, it looks sorta comic like, but the tech still needs work OR the process needs FAR more time and energy to find those moments. The creators using the program, also needs an education in visual storytelling to know what is needed as well. Maybe draw crude sketches of the visuals needed and have the program flesh it out? That sort of thing. I'm not saying it may not be able to be done. But this isn't quite it.
I agree with you. The point of the video was to demonstrate how you can put something like this together relatively quickly. By no means am I trying to say that this is enough or that one shouldn't put in more effort. As you might imagine, producing video content takes enough time as it is. So consider this a "demo", not the final product.
@@TokenizedAI Right, as a content creator myself, I completely understand. Ideally however, it would be good, perhaps next time, to let amateurs know what may be involved in visual storytelling that isn't being presented. Like emotional beats, screen direction, establishing shots, Master shots, etc. I've watched so many videos similar to this, seeing if anyone has "cracked the Ai making comics thing," and have yet to see it. The videos are very superficial (make images that look comic like). It's not like I "want to be replaced," but I'm really curious if it's even really possible. And I'm not willing to put the time and effort myself because it's just way easier for me to draw what I want then to fight a program into giving me what I can easily do myself. It actually takes MORE work to have Ai do it. Also, I'm not willing to pay for Midjourney, I use Stable Diffusion, since it's free. SD doesn't do drawings well at all though (except anime or manga style, which is very limited since it's only one style. And I've downloaded and tested a LOT of models).
Honestly, I don't think anything has been cracked or can be cracked, as you say it. People are focusing way too much on the superficial aspects of the process. The artists are upset because they think they'll become obsolete. And others think they can make awesome comics just because they have images now. AI makes A LOT of things easier and more accessible. Especially to those who may have almost all the skills, except for 1 puzzle piece. But in the end, the only people who will ve able to produce anything meaningful are those that are creative to begin with. They may not be illustrators but they have creativity inside of them. And artist or storytellers are just going to work very differently. Ideally, they will become more productive, faster, get more attention. In the end, the key thing to understand is that EVERYONE needs to work on distribution (read: an audience). Because that's the one thing that isn't easy to build or costs a lot of money. Doesn't matter how good someone is if no one knows who you are.
@@TokenizedAI So, because of this conversation, I decided to try to make a comic using SD. I recorded the whole thing and will eventually post it on my channel. So I did a few things. I did a three panel gag strip. That way it reduced the complexity. It had a setup, middle, and punchline. Then I used SD as an assistant. I did my regular comic process of thumbnailing and setting up the shots roughly using Clip Studio Paint (CSP). I then had SD generate a background (BG) from scratch for the shot. Once I got something I could work with I imported in into CSP and roughly fleshed out the BG a bit more myself. I then gave the BG back to SD to redo using inpaint. I asked SD to export a high resolution BG, took in back to CSP and added rough characters on it, on separate levels. I took the rough character drawings individually (separated from the BG) and had SD flesh them out using inpaint. Once I had something useable, exported high resolution versions, imported them into CSP, put the characters on the BG, and fixed anything wrong with them myself. Redrawing hands, mouths, eyes, that weren't working or were not telling the story properly. Once I had characters, I made copies of them and added them to the next panels. I modified the characters roughly in the last two panels along with the BG in order to get the shots I wanted, and had SD go over my modification to tie them up a bit using inpaint. The whole time keeping in mind proper screen direction and the needs of the simple strip. The result was good. BUT I did as much work as the Ai. I guided and collaborated with the Ai, it didn't do everything. The result is something I'm not entirely sure I can take any real credit for. It doesn't look like anything I'd ever draw. It's not in my style. But it's got my visual storytelling and my compositions. I've be interested in trying it again, with another gag strip with a tad more complexity.
Automation always removes basic tasks but tends to add a lot of others, that usually require more skill. So I don't see immedate risks to specialists. However, it does become a problem when content production is so great (i.e. quantity) that supply outpaces demand. The latter has a natural cap, the former does not.
Christian,, this is fantastic work! Whoever has tried to do this before (I certainly have spent tons of hours trying to get it right), knows that what you did here - despite using AI tools - still requires a lot of creative thinking, tweaking and LOTS of time! Kudos for sharing it with us. You helped me to get a few things right with the Graphic novel I am working on. Thanks for sharing!!!
How to generate 3 or more multiple consistent characters interacting with each other and use them in different scenes? And how to add people behind them??? It's impossible???
This was the BEST video I have seen on AI! I love that you not only go step-by-step, but that you go slow enough to be understood/followed. I love that you are very explicit with your prompts. You have a new subscriber and I look forward to watching your videos.
Hey there! I just watched your video and I have to say, I'm blown away by your creativity and talent. It's amazing to see how you've used Midjourney to bring your comic book to life. The story and artwork are fantastic, and I can see this becoming a hit with readers. I appreciate you sharing your process and insights with us, it's very helpful for those of us who are looking to create our own comics or graphic novels. Thanks for sharing your amazing work with us and keep up the great work!
Thanks for the tutorial. This video makes me even more excited to explore more AI capabilities to help my work. I have tried other AI based Text to Image Generators like 'Leonardo AI' which is pretty good too. But in general I must admit that the image produced by Midjourney is much closer to what we expect. Too bad Midjourney no longer allows trial users to create images at all. Actually it would be better if they didn't ban it at all, but just limited the amount per day, and reset it for the next day, like Leonardo did. Thus, people no longer need to create a new user every day.
a cutoff date for ChatGPT (GPT4) training data is September 2021. There was no midjourney yet. Then the information for GPT4 that the panels will be filled with midjourney output doesn't make any impact, unless chatGPT has been trained by users (us) and retained this training in live database. I am puzzled...
GPT-4 is mostly hallucinating here, but Midjourney existed before the cutoff date. However, if you wanted to, you could feed in a bunch of information at the start of the chat thread that explains what MJ is. I didn't do that here because it wasn't necessary.
No, I used Affinity Publisher. You don't need a special comic tool to do this. Any layout software will do this with some tweaks. I even share the Publisher files in my Supporting Material in the description.
I can create my own story, but I just need an AI tool that will give me a lot control over the images I create. I need character consistency, posing, ability to change the camera angle, background consistency, option to change the environment of a scene (e.g., time of day, lighting conditions, weather, season), and ways to create accurate scenes according to my instruction. I hope more features will come that give us greater control over these visual aspects. Even if this amount of control isn’t possible right now, it might become possible soon, judging by the speed that AI has been progressing. Personally, I think character consistency is at the top of the list when creating a comic, because it would look too confusing if characters change appearance from scene to scene.
All of this can be achieved as long as you're willing to learn to use various different tools. As much as I understand that people want the "one tool to rule them all", it's a bit of an unrealistic expectation. Nothing of value is ever easy. And once it becomes easy, it immediately loses its value. Food for thought.
So so COOL! Really good concept! Got to try! One doubt thats bothering me is that, aren’t all these generated images on the Midjourney feed for all to see and download? Or is there a way to hide or keep it private?
It's ironic or even hypocritical that you want to create content off the work of other people but then you want to have rights to it. Welcome to you will own nothing an love it!
Thanks a lot, this was a very, very helpful video!! I am pretty new with Midjourney, but also faced the big problem with consistancy. I wonder, why there is no prompt or option in Midjourney itself, to "freeze" a face or person as being consistant for future use. This need is so obvious, but seems not to be solved yet!
Great video ! I've got a question, you adressed how consistency in character design is not respected on purpose and you said should you use the proper functions it'd not be the same. My question is how much consistency can you keep using the proper methods ? (in terms of percentage let's say). I am asking this because that could be a really good way to make concept art in order to pitch my story to an artist.
I would simply do a lot more photobashing. E.g. create the scenes with little respect to the faces, then smack the face I actually want onto the bodies and then rerun the same prompt of the scene with the photobashed image reference. Still not perfect but much better. It just takes to long for this video format.
I very rarely subscribe to A.I. channels, but this one is absolutely gold. Thanks for the in-depth tutorial, it's easy to understand, and highly informative. Insta-subbed and look forward to more of your videos in the future!
In real life: Between sceneries, tools and outfits, even general appearance, can change wildly. This is why Midjourney variation can work well even for comic book format.
You are so pleasant to listen to, the tone of your voice is calming and the speed at which you speak is not too slow, or fast and you are very clear, thank you!
Thanks. I appreciate how you talk us through your process. I would love to hear about your negative prompting technic since the --no parameter seems to be broken in 5.1 and 5.2
Thank you so much for posting this process, i have been trying to crack this nut for some time...creating a comicbook but not being an artist....when i learned about midjourney I begun investigating...your information moves my venture alittle closer to being a reality for me...artists are safe you still need to finish drawing the characters to keep consistant throughout your story and brand....human artists will be around for alittle while longer...for sure..
Your best bet is to hire an artist. There's a lot of hate towards AI in the comic community, I imagine it would be difficult to gain support/following. Best of luck on your endeavors.
@@roymakescomics thanks for the comments, no money for artist...which is probably a common issue for non artists...i agree that AI art is not ready for prime time...but it is a good starting point...
Hi christian, thanks for all that you do. I check daily on your channel in hopes of new content. Email notifications are awesome too - Thank you! A quick question, my created character is now very consistent due to the data base of images I have created but I am still struggling with getting MJ to create an image with him back on to the camera and looking in to the distance. Any advice for the prompt?
Are you on the mailing list? Make sure you are signed up for that. That's where I announce it first. I won't do a video on it until the full public release. You can find the link to it in the description text of all my videos.
As always, a great video Christian ! I have been working on something to your project and also wanted a certain "look" or "Style" for the characters and background . To get something similar to what I was looking for, I took an image with a look/style that was similar to what I envisioned, and used the " /describe " function in MidJourney to provide a prompt style description that I could use. I then adapted that style description for my needs. Sort of like reverse-engineering the style and look from something to fit my needs. The results were kind of close but still needed a lot of re-rolling ! 😀
Re-rolling is underrated. I think a lot of people believe that it's a sign of bad prompts but it's not. We underestimate how much randomness plays a role in all of this.
This is just fantastic, thank you so much for taking us through this and helping us. Do you have a guide or video that shows us how to put a title onto an image
For all the people complaining, sit down. This is a tool for Artist, this will not end the Artist era that is completely impossible I am a Artist myself, this makes it much easier to define and make story telling way better for your comics, novels, anime.
Nice joke, China game company already fired many game illustrators :) This is not a tool, its to slowly replace digital artists. Traditional artists are safe for now but nothing digital will ever be safe and 3D will be next. These kind of jobs will be hired even less along with design cause instead of needing 6 workers, now 1 worker can do all of them with the help of this new "tool". Writers themselves won't need to hire any illustrators cause AI exists
A big problem on the comics business is that there’s very few readers. To me that would be the attack angle, can we create AI readers? Any tips on getting AI to read comics? (Serious question)
@@TokenizedAI That’s the thing though. No readers no money. Money is digital, so why shouldn’t we develop AI that are consumers? What’s the reason not to? Sounds like a better idea than developing AI that makes stuff there’s no demand for. But money is always in demand.
A trick I use for creating consistent characters was to base them on real actors. It works best when it only looks70-80 percent like the actor so there's not a copyright problem.
I can do character profiles and draw them. I can can do scenes and the overall plot. Dialogue, fleshed out story, character development for everyone not that main character and why everyone gets along or not in a way where they don’t just feel like scripted npcs and all of that is beyond me.
@@TokenizedAI All of this time I’ve been caught up and stalling because I can’t do the heart of the story, the character interactions and dialogues, the journey as you gave as the premise in your example. You get that right, and you almost don’t even need the elements that I can do. But, you just made me realize that a drama / opera isn’t the only way to tell a story. You can take a narrative approach. Like a journal entry or summarization of the days events or life story and keep the in-depth character details and lore separate.
Perché non sono un pessimista così incredibile come te che apparentemente ha bisogno di trovare qualcosa di negativo in ogni cosa. Perché non ti godi la vita così com'è? 🤷🏻♂️
How do you define a consistent building or interior? When do you consider it consistent and where's the point at which you longer consider it consistent? And why is the consistency important for your use case? Is consistency really that important for what you want to do? Your prompt defines a bunch of inputs that the model then tries to approximate via a diffusion process. It's not a 3-dimensional blueprint of a building with furniture placed in certain locations and shots taken from different a angles.
I am so happy I came across your channel recently. Ive been up all night, ive subscribed and downloaded the links awesome stuff. Im about to open the zip file from comic book, I feel like Ive just been given something really special and totally free. i especially love the midjourney cheat sheet ive been looking for something like this. You deserve a Medal.... :) Thank you so much.
Thumbs up once again. The best tips on comics and midjourney. The most valuable tip for me was that the characters don't have to be 100% coherent. I took a look at some of my old comic books and lo and behold, the cartoonists didn't exactly cover themselves in glory when it came to consistency back in the day (in the golden age of comics in the 50s). A blonde in a red dress and a brunette in a grey costume met a redhead in a grey suit and a black-haired man with a moustache in a blue mechanic's suit. So you have to differentiate the characters well, then it should work. I started my own project tonight (a Victorian steampunk comic)....
A small idea for a video, the new permutations should be doable without pro, since it uses no new feature, I think a simple Discordbot should do the same trick.
Hello, Christian. As usual, a polished and sleek video with all the necessary documentation. Thank you very much; I've been following you for quite some time and am continually learning new things from your videos.
What if I want to create consistent SCENES/ENVIRONMENTS rather than just characters? E.g. a character's bedroom that he returns to at different points in the story. How can I ensure I can get the bedroom to look the same or similar? Would I follow the same guide as consistent characters i.e. generate scene images and then paste in multiple URLs? Or is this impossible? Thank you
Two problems: 1. Everyone means something different when they say "consistent environment". 2. You can't really do that because MJ isn't actually aware of space. It doesn't know it's a room. So it can't show you the same room from different angles. It can show you a similar room from another angle, but not the same. If you use image references, you'll be able to transfer the style of the image and to a certain extent the interiors, but it will always be a different room.
@@TokenizedAI There is a way forward, but you'll need a 3d program like Blender or Daz Studio. Arrange a room scene, render it, then send the image to MJ. This way you can get the same room from within a cube space, although AI may make subtle changes to the 3d render depending on your prompt.
Great man! I think you know this already but IF we wanted more consistency (which in this case is NOT a problem as they look consistent enough) there is this old trick of using a celebrity BUT changing his/her gender to opposite or change his/her age or even style like anime style etc. That does not solve the different clothes problem 100% but maybe there is a solution for that too. Also using already created images as an image source could be helpful. Many thanks! Looking forward for those files. Great video!
I’ve never used midjourney yet personally but I think you are right. If you used image reference prompts like for the character and background I think that would help with the consistency issue. Would be a lot more work but nothing compared to drawing everything by hand lol. I think midjourney as whole has a language and process that’s yet to be fully discovered. I truly believe you could make a full graphic novel using this but just needs the correct prompts and processes in the correct way to make it possible.
This was extremely helpful, and I really appreciate that. You were willing to take the time to share all of this with people like me who love this kind of stuff and love this kind of use, comic book, usage of AI. Thank you, and will certainly be providing my email address for, these amazing prompts.
Nice video Christian. Feedback: it seemed you were angry about the feedback you receive about the "consistent character" issue. Instead, make a joke of it or tease and have fun. Love your videos, thank you.
I wasn't angry. It also wasn't feedback that I received. I was challenging the people with unrealistic expectations, who expect Midjourney to be able to do literally everything under the sun, when only 5 months ago none of them were able to do any of this because the tech wasn't even available. People are too impatient and I think it's more than fair to say so.
Amazing work-thanks for this! Do you think MidJourney and other text-to-image tech will eventually be able to make consistent characters an integrated feature? To be able to save a detailed description/look of a character and have the AI put that same person in whatever scene/style you choose seems like it should be in reach, where it can do it pretty decently with famous people. But I lack the understanding of how it works..maybe I’m just daydreaming :)
Yes, but not in the way you might think. O believe MJ wants to eventually allow people to create their own mini-models for consistency. But I think that's still far away.
Super Video, danke. War sehr aufschlussreich. Beim Dialog hat GPT-4 aber noch Verbesserungspotential: Der Dialog startet immer mit Maya und die Antwort ist meist etwas einsilbig. Das bringt nicht gerade ein spannendes Story-Feeling auf. Aber so als Baseline und Inspiration schon mal nicht schlecht. Vielleicht schafft es Midjourney in einer der nächsten Versionen sogar, konsistente Charaktere zu generieren (ohne, dass man Midjourney zuerst mit ihnen trainieren muss). Dann könnte man sich, statt fernzusehen, UA-cam zu schauen oder ein Comic-Heft zu lesen, ein eigenes Comic-Heft erstellen, von dem man die Geschichte noch gar nicht kennt. Sicher witzig. So ähnlich, wie sich selber eine Gute-Nacht-Geschichte erfinden/erzählen. Toller Kanal. Weiter so. Viele Ostergrüsse aus der Schweiz 🇨🇭
Ja, der Dialog ist etwas hölzern und Leo kommt etwas "dumm" rüber 😅 Allerdings muss ich für die Videos alles ein Stückweit knapper halten. Es geht ja primär darum, ein Proof of Concept zu liefern 🙂
Hey, I'm also trying to grasp the idea of having multiple characters in one setting but I've been at it for like 2-3h and I can't seem to get any results, I get only ONE character every time. The prompt I've used was this: /imagine prompt: Blend of comic book art and lineart in full natural colors. In a bustling medieval city, a wise, middle-aged paladin with a white beard and long, flowing hair who wears ornate silver-and-gold armor and a blue cloak and a young paladin in his early twenties with long golden hair who wears royal blue-and-gold armor with a matching cape meet each other at the city gates --v 5 --ar 9:4 --s 750 --v 5 --s 750 --v 5 --s 750. Am I doing something wrong? I've been following your guide pretty much step by step.
Why are you adding the same parameters 3-4 times? I really recommend you watch some of my videos on syntax first. I also have a video on multiple characters.
@@TokenizedAI sorry, I think it was just a copy paste issue (i’ve edited the post a few times to make it look solid but I didn’t notice I’ve added the parameters multiple times 😩 I’ve seen your multiple character video aswell, it was actually the video that inspired me to go on this journey. (I work as a surgeon but I’ve always wanted every since I was a kid to create my own comic books… so thank you for all the great info!)
Great video, my current project is actually taking3 short stories I've already written and trying to use MJ to create not just a better front and rear jacket but create a read along either comic or some other medium, obviously this is a lot to learn but your video is literally what I'm attempting right now, I am also following on Twitter so I will comment there at times also maybe with more Q's
What exactly is your point? Like... seriously. When was the last time you actually enforced a copyright on your creation? 90% of what the average user creates gains absolutely nothing from having copyright. This obsession with copyright is so strange. Especially among people for whom it's essentially meaningless beyond personal preference. Unless you've created intellectual property that has widespread recognition and brand value, you couldn't care less about having copyright protection. Here's another example: If you create a comic book and outsource the illustration to an artist, you also do not automatically hold copyright on those images. You get usage rights, unless of course you coerced the illustrator into a raw deal. In some countries you can't even transfer copyright at all. But you can transfer usage rights. And you do hold copyright on the overall comic book creation. So please enlighten me, how is this copyright discussion relevant in 90% of the practical cases beyond it being a meme at this point? It's an academic discussion amongst lawyers at best.
@@TokenizedAI Im Not talking about legalities , in talking about self publishing companies cracking down on AI . They are not accepting anything that’s been AI generated . Don’t lie to these folks who think they’re gonna make thousands of dollars .
I watched this video. Good prompts, great info. But I have a few suggestions, to create consistent characters, I'd create them in various poses on white background first, then remove the background and save as png files. Then I would create separate backgrounds of forest, river etc and place these character png files in the foreground and create dialogue boxes to have consistent characters throughout. It takes a bit of work, but it is worth it in the end. You'd have the same characters throughout making for an effective comic book.
I'm aware of how to do it and also believe I even said so in the video. I have an entire series on character design on this channel. This was just a demo and by no means meant to imply this is the whole process. As you say yourself. It takes work. Hence why it's far too much to stuff into a YT video. People tap out if the video is much longer than 15 minutes.
Thank u for your content, it's amazing! Soooo 😂 you said "go learn how to draw..." And I actually did it many years ago, and I can draw, I'm a professional illustrator, and I'm watching your amazing tutorials and do some experiments in Midjorney but I don't understand how you create a comic book for one day? I spent a few weeks just to get illustrations I wanted and they still need to be fixed in Photoshop. Ofcourse if I had to draw them by hands that would take me a month.... But still! I feel like my process is way too slowly that it could be...
If you pay attention to the images in the video, you'll see that the characters aren't 100% consistent. This is the compromise that I made and that's why I completed it in a day. If wanted everything to be perfect and also a better story, then of course, it would have taken much longer.
Great video. Unfortunately US law states that intellectual property can be copyrighted only if it was the product of human creativity, and the USCO only acknowledges work authored by humans at present. Machines and generative AI algorithms, therefore, cannot be authors, and their outputs are not copyrightable. (Hence, if you don’t mind your work being in the public domain, go for it.)
So you tell your audience that character consistency is IRRELEVANT in this video, after 16 minutes? Thanks.
Where did the thumbnail, title or intro promise you a lesson on character consistency? Exactly, it didn't. You're projecting your assumptions and blaming me for that. How does that make sense? Right, it doesn't.
My "audience" knows that I have an entire playlist on character design and the end screen of this video even promotes it. You, the random viewer may not know this, but then you're technically not really my audience.
So I suggest you direct your frustrations elsewhere or inward. You're getting a free education here, so I recommend dialing down the sense of entitlement. It clearly isn't doing anything for your mental health.
This video has a like ratio far beyond 99%, so clearly the video isn't the problem. Yes, that's a hint for you...
Buuuuuuuuurrrrrrnnnnnnn🔥
Christian, you are extremely generous with your information and resource links. Thank you.
Just paying it forward 🙂
Looking for consistent characters?
Big fan of the channel, Christian. We've used your videos for training staff. I suspect a lot of people will get hung up on this even though you explained it a bunch of times...haha....
If you're new, go see the earlier videos on character building, we're doing this stuff professionally (I have a small staff doing character creations for video shorts) and this channel has been our go-to channel for midjourney stuff.
That said, the "I want a quick tip" verison can be summed up like this:
In the first step, don't use any background (I think Christian did this to make a point, but it's actually more work and can create other problems later). Creating this character on a plain background also has the added advantage that you can remove the background and use the character on a transparency layer to add to a scene (like the one he wanted 3 characters but knew that MJ couldn't do that correctly).
Next, once you have a character or character you like, save the URLs for that character (whether it's 2 or 15 URLs, just save the short links).
Then, instead of prompting like this: "An asian woman in her blah, blah, blah, blah" over and over, just prompt: A woman URL1 URL2 URL3 URL4 etc, (as many as you need) in SITUATION and SCENE ( in a boat on an Amazonian river or in an Asian jungle or fighting a snake or holding a map or entering a cave, whatever....) IN THE STYLE OF and then whatever you're using here to make the style you want (you can cut and paste this instruction over and over for a consistent style, whatever that is).
And there you go...
As a final note: You can make adjustments to the style and chaos while leaving the URLs and the SITUATION and SCENE and IN THE STYLE OF instructions identical and get much different results, so it's worth playing with those too (not just rerolling).
One other thing I've noticed, MJ does seem to make changes in style and look when you use a different aspect ratio. So, for example, sometimes I'll use --ar 2:1 even though I want a square panel because the look is better for that scene and I just trim off the sides, so don't forget that you don't have to make the aspect ratio in MJ the aspect ratio you need for a panel or page or scene, you can make it different to get the look you want and then cut it out in PS or GIMP or whatever to the right size. Sometimes this is less work than trying to force MJ to create a certain image in a certain aspect ratio. Imho.
Thanks again, Christian, you've been a monster-sized help to us.
This won't create 100% consistent characters either. Just worth adding
Depends on what you mean by 100%.
Of course, the more stylized it is, the more variations you'll get.
And I'm not claiming every single iteration is consistent, but if you run the prompt a few times with the same URLs (the number of them depends on what you're doing) you'll get a consistent character 100% of the time you're willing to put in the work to produce a consistent character.
Thank you for these tips. Really useful.
@@TanChoonHong I'm finding that the situation will change the character when the training model's images can't figure out what to do, so, for example, using this model to put a character on a horse doesn't seem to work very well.....it knows the character, and it stays consistent, and it knows the word "horse" and will even put a horse in the image, but "riding" a horse...I can't quite get there..
That said, I'm getting very consistent images by using a group of URLs that force the AI to stay in a very narrow range of possibilities and sometimes it will put them where I want them.
A lot of what I'm doing requires me to make a consistent character that is happy or sad or sleepy or whatever on a plain/white/transparent background and then layer that over the scene I want.
If you try to use the method I mentioned above, well, I sort of jumped the gun on what it can really do, which is why, I suspect Christian didn't even bother with it, it's too hard when the scenes are complex to get everything right, better to make multiple images, and combine them in PS or GIMP or whatever.
Good luck!
@@michaelsbeverly Agree, faster to generate on white/transparent background & composite with other images. Thanks.
"just go learn to draw" 🤣 Glad someone finally said it!
No point in beating around the bush 😅
@@TokenizedAI I actually tried to "just go learn to draw". But I just suck at it. Best way to get "'Exactly" what I want would be to draw it myself. Still sucks when using a tool that's 1000% better at drawing that still takes 30+ prompts to "Almost" get what I ask for. Feels like my brain is on fire when I'm constantly changing and modifying my prompts but getting anything but what I need. Plus I've found that 90% of the videos on this is just having A.I do 100% of the work. Is it really bad that I want to use A.I to create a comic book for "My Stories" because I am truly horrible at drawing. In short, I have tons of stories to tell. Just no talent in drawing or thousands of dollars to hire artist. I wanted to use gofundme, but would be impossible without something to show process. Can't even start without some sort of funding, but don't think i would get far on promises alone. Considering all the scams out there.
omg !! i was just watching your seed video !!!! im sp excited that this was just posted !!
Hope you enjoy it 🙂
The art of visual story telling is more than just pasting a bunch of random images together, comic artists jobs are safe, for now.
That's true. And I said it before and I will say it again. This was only meant as a demo of how these tools can be leveraged. It wasn't meant to be a final product.
The consistency in the characters images are maybe possible with Insight FaceSwap ... creating first the stories images with Midjourney and then swapping faces .... I tested it rudimentery and it looks like a possibility. If you are good in storyboard and want to stand out of the drawing crowd it could be a possibility. Anyhow ... I think in a year we will talk advanced situations with this ... 🙂
Just a matter of months sadly
I tend to agree with Cloudbreaker on this one.
@@TokenizedAI is there a way to do this off your own sketch ups and lay ins, story boards etc?
Christian, you have given me a breath of fresh ideas and concepts that will enhance my journey in making good stories and images. I have been looking for an educator such as yourself, and I thank you!
one thing that you could do is draw the pictures the way you want them in a bad/rough way but with all the right composition and upload them as references for midjourney to create around. would probably turn out really good
Fantastic work on the comic book! I am thrilled to see how you employed Midjourney to evoke 'Treasures Beyond Gold' with distinctive imagery.
Thank you for the lead. I've been trying to make consistent characters in v5 and the only thing that has got me 90-95% close is pasting the url of the target image and then following it with the descriptive prompt, only you have to maintain the identity names of the character/s you are generating.
I found your video interesting. The visuals look very impressive and look like comic art. I find that fascinating. However, I draw comics and I create storyboards professionally, there's more to comics than just "characters standing/sitting/looking" in an environment. There's acting, and pacing. There's....well...visual storytelling that needs to happen. A good test is if you can look through the visuals of a comic without the dialogue or the narration and still get the 80% of the story. Where were the puzzle solving shots? The moments of friendship? The visual collaboration? The emotional beats? The action moments? The danger moments? The resolution moments. All I saw where images of people in midshots mostly looking at the middle distance. I also didn't see an example of the final with dialogue balloons.
My point is, yes, it looks sorta comic like, but the tech still needs work OR the process needs FAR more time and energy to find those moments. The creators using the program, also needs an education in visual storytelling to know what is needed as well. Maybe draw crude sketches of the visuals needed and have the program flesh it out? That sort of thing. I'm not saying it may not be able to be done. But this isn't quite it.
I agree with you. The point of the video was to demonstrate how you can put something like this together relatively quickly. By no means am I trying to say that this is enough or that one shouldn't put in more effort. As you might imagine, producing video content takes enough time as it is. So consider this a "demo", not the final product.
@@TokenizedAI Right, as a content creator myself, I completely understand. Ideally however, it would be good, perhaps next time, to let amateurs know what may be involved in visual storytelling that isn't being presented. Like emotional beats, screen direction, establishing shots, Master shots, etc. I've watched so many videos similar to this, seeing if anyone has "cracked the Ai making comics thing," and have yet to see it. The videos are very superficial (make images that look comic like). It's not like I "want to be replaced," but I'm really curious if it's even really possible. And I'm not willing to put the time and effort myself because it's just way easier for me to draw what I want then to fight a program into giving me what I can easily do myself. It actually takes MORE work to have Ai do it. Also, I'm not willing to pay for Midjourney, I use Stable Diffusion, since it's free. SD doesn't do drawings well at all though (except anime or manga style, which is very limited since it's only one style. And I've downloaded and tested a LOT of models).
Honestly, I don't think anything has been cracked or can be cracked, as you say it. People are focusing way too much on the superficial aspects of the process.
The artists are upset because they think they'll become obsolete. And others think they can make awesome comics just because they have images now.
AI makes A LOT of things easier and more accessible. Especially to those who may have almost all the skills, except for 1 puzzle piece.
But in the end, the only people who will ve able to produce anything meaningful are those that are creative to begin with. They may not be illustrators but they have creativity inside of them.
And artist or storytellers are just going to work very differently. Ideally, they will become more productive, faster, get more attention.
In the end, the key thing to understand is that EVERYONE needs to work on distribution (read: an audience). Because that's the one thing that isn't easy to build or costs a lot of money. Doesn't matter how good someone is if no one knows who you are.
@@TokenizedAI So, because of this conversation, I decided to try to make a comic using SD. I recorded the whole thing and will eventually post it on my channel.
So I did a few things. I did a three panel gag strip. That way it reduced the complexity. It had a setup, middle, and punchline.
Then I used SD as an assistant. I did my regular comic process of thumbnailing and setting up the shots roughly using Clip Studio Paint (CSP). I then had SD generate a background (BG) from scratch for the shot. Once I got something I could work with I imported in into CSP and roughly fleshed out the BG a bit more myself. I then gave the BG back to SD to redo using inpaint. I asked SD to export a high resolution BG, took in back to CSP and added rough characters on it, on separate levels.
I took the rough character drawings individually (separated from the BG) and had SD flesh them out using inpaint. Once I had something useable, exported high resolution versions, imported them into CSP, put the characters on the BG, and fixed anything wrong with them myself. Redrawing hands, mouths, eyes, that weren't working or were not telling the story properly.
Once I had characters, I made copies of them and added them to the next panels. I modified the characters roughly in the last two panels along with the BG in order to get the shots I wanted, and had SD go over my modification to tie them up a bit using inpaint. The whole time keeping in mind proper screen direction and the needs of the simple strip.
The result was good. BUT I did as much work as the Ai. I guided and collaborated with the Ai, it didn't do everything. The result is something I'm not entirely sure I can take any real credit for. It doesn't look like anything I'd ever draw. It's not in my style. But it's got my visual storytelling and my compositions. I've be interested in trying it again, with another gag strip with a tad more complexity.
Automation always removes basic tasks but tends to add a lot of others, that usually require more skill. So I don't see immedate risks to specialists.
However, it does become a problem when content production is so great (i.e. quantity) that supply outpaces demand. The latter has a natural cap, the former does not.
Great video! You inspired me to do my own comic book, and mine turned out amazing! Thanks so much!
Christian,, this is fantastic work! Whoever has tried to do this before (I certainly have spent tons of hours trying to get it right), knows that what you did here - despite using AI tools - still requires a lot of creative thinking, tweaking and LOTS of time! Kudos for sharing it with us. You helped me to get a few things right with the Graphic novel I am working on. Thanks for sharing!!!
It's nice to see that some people appreciate the amount of work this takes 🙂
How to generate 3 or more multiple consistent characters interacting with each other and use them in different scenes? And how to add people behind them??? It's impossible???
Wow, I hadn't toyed around with comicbook style drawings. These look amazing!
Amazing lesson Christian. Thank you. Your generosity knows no bounds!
You're welcome 🙏🏻
This was the BEST video I have seen on AI! I love that you not only go step-by-step, but that you go slow enough to be understood/followed. I love that you are very explicit with your prompts. You have a new subscriber and I look forward to watching your videos.
Appreciate it 🙏🏻
Hey there! I just watched your video and I have to say, I'm blown away by your creativity and talent. It's amazing to see how you've used Midjourney to bring your comic book to life. The story and artwork are fantastic, and I can see this becoming a hit with readers. I appreciate you sharing your process and insights with us, it's very helpful for those of us who are looking to create our own comics or graphic novels. Thanks for sharing your amazing work with us and keep up the great work!
Just trying to help give people ideas 🙂
Very interesting! And I like the step-by-step-approach that You take! This is helpful.
Pleasure 🙂
Wow, Christian - this video was GOLD!
Thanks 😀
Thanks for the tutorial. This video makes me even more excited to explore more AI capabilities to help my work.
I have tried other AI based Text to Image Generators like 'Leonardo AI' which is pretty good too. But in general I must admit that the image produced by Midjourney is much closer to what we expect.
Too bad Midjourney no longer allows trial users to create images at all. Actually it would be better if they didn't ban it at all, but just limited the amount per day, and reset it for the next day, like Leonardo did. Thus, people no longer need to create a new user every day.
Yeah they had to shut down the trial memberships due to excessive abuse 🫤
Thank you for this amazing video! i am wondering though, could this be done with more than 2 consistant characters? lets say 4 or 5?
You are the MJ prompt master bro!
Seriously a prompt god
Not sure about "Master" but I have my moments 😁
a cutoff date for ChatGPT (GPT4) training data is September 2021. There was no midjourney yet. Then the information for GPT4 that the panels will be filled with midjourney output doesn't make any impact, unless chatGPT has been trained by users (us) and retained this training in live database. I am puzzled...
GPT-4 is mostly hallucinating here, but Midjourney existed before the cutoff date. However, if you wanted to, you could feed in a bunch of information at the start of the chat thread that explains what MJ is. I didn't do that here because it wasn't necessary.
Thanks! How come you didn't use Comic Life 3 to make the cover page?
Why should I have?
@@TokenizedAI just curious if Comic Life does not have that ability?
I honestly don't even know what Comic Life is 😬🙈
@@TokenizedAI that is what you used in this video to arrange your comic, didnt you?
No, I used Affinity Publisher. You don't need a special comic tool to do this. Any layout software will do this with some tweaks. I even share the Publisher files in my Supporting Material in the description.
Thanks for sharing. Will try myself tomorrow!
Go for it!
Amazing work brother!
Great video ! always awesome content on your channel btw 👏
Thank you so much 😀
Wow - das ist richtig spannend. Und du hast einen sehr guten Überblick gegeben. Die Möglichkeiten damit sind verrückt. Echt super. Danke
This was really helpful thanks a lot. Is it also possible to animate pictures lightly?
Kind of, but it's easier with photo.
I can create my own story, but I just need an AI tool that will give me a lot control over the images I create. I need character consistency, posing, ability to change the camera angle, background consistency, option to change the environment of a scene (e.g., time of day, lighting conditions, weather, season), and ways to create accurate scenes according to my instruction. I hope more features will come that give us greater control over these visual aspects. Even if this amount of control isn’t possible right now, it might become possible soon, judging by the speed that AI has been progressing. Personally, I think character consistency is at the top of the list when creating a comic, because it would look too confusing if characters change appearance from scene to scene.
All of this can be achieved as long as you're willing to learn to use various different tools. As much as I understand that people want the "one tool to rule them all", it's a bit of an unrealistic expectation.
Nothing of value is ever easy. And once it becomes easy, it immediately loses its value. Food for thought.
Good and nice . Now i am confused with the part where to put the photos . Publisher?
Whatever floats your boat. Publisher would work just fine.
So so COOL! Really good concept! Got to try! One doubt thats bothering me is that, aren’t all these generated images on the Midjourney feed for all to see and download? Or is there a way to hide or keep it private?
Yeah, if 100% privacy is your thing then you need to go for the Pro plan. MJ's philosophy is very open and about sharing.
It's ironic or even hypocritical that you want to create content off the work of other people but then you want to have rights to it. Welcome to you will own nothing an love it!
Thanks a lot, this was a very, very helpful video!!
I am pretty new with Midjourney, but also faced the big problem with consistancy. I wonder, why there is no prompt or option in Midjourney itself, to "freeze" a face or person as being consistant for future use. This need is so obvious, but seems not to be solved yet!
There's no option because this technology is far more complicated under the hood that most people seem to think 😆
And how do we add the dialogue? No link for affinity...
I'm not sure I understand your question.
How do we add the speech or thought bubbles
In addition to the pictures, the characters are speaking. How do you insert speech? In a speech bubble or below, above or where?
Did you get character paintings in midjourney or affinity...
Midjourney
Great video !
I've got a question, you adressed how consistency in character design is not respected on purpose and you said should you use the proper functions it'd not be the same. My question is how much consistency can you keep using the proper methods ? (in terms of percentage let's say).
I am asking this because that could be a really good way to make concept art in order to pitch my story to an artist.
I would simply do a lot more photobashing. E.g. create the scenes with little respect to the faces, then smack the face I actually want onto the bodies and then rerun the same prompt of the scene with the photobashed image reference. Still not perfect but much better. It just takes to long for this video format.
I very rarely subscribe to A.I. channels, but this one is absolutely gold. Thanks for the in-depth tutorial, it's easy to understand, and highly informative. Insta-subbed and look forward to more of your videos in the future!
Thank you, I really appreciate that. 🙂
But in order to dot his, we have to PAY Midjourney first, right? I mean, we can't pull it off using the free version, right?
There is no free version anymore.
In real life: Between sceneries, tools and outfits, even general appearance, can change wildly. This is why Midjourney variation can work well even for comic book format.
I like this channel and the attitude of the presenter and how his approaching the explanation. I've subscribed
Thanks man! 🙂
I need to storyboard a movie script. In your opinion, what product do you think is best?
What exact "products" are you talking about?
You are so pleasant to listen to, the tone of your voice is calming and the speed at which you speak is not too slow, or fast and you are very clear, thank you!
Apparently it's the perfect speed. Slow enough for non native speakers and when you switch to 1.5x, it still works 😅
@@TokenizedAI yesterday I watched an American AI UA-camr and the whole video I kept thinking, why are you yelling? 😆
Thanks. I appreciate how you talk us through your process. I would love to hear about your negative prompting technic since the --no parameter seems to be broken in 5.1 and 5.2
I have a separate video on this in my channel.
Thank you so much for posting this process, i have been trying to crack this nut for some time...creating a comicbook but not being an artist....when i learned about midjourney I begun investigating...your information moves my venture alittle closer to being a reality for me...artists are safe you still need to finish drawing the characters to keep consistant throughout your story and brand....human artists will be around for alittle while longer...for sure..
Your best bet is to hire an artist. There's a lot of hate towards AI in the comic community, I imagine it would be difficult to gain support/following. Best of luck on your endeavors.
@@roymakescomics thanks for the comments, no money for artist...which is probably a common issue for non artists...i agree that AI art is not ready for prime time...but it is a good starting point...
Hi christian, thanks for all that you do. I check daily on your channel in hopes of new content. Email notifications are awesome too - Thank you! A quick question, my created character is now very consistent due to the data base of images I have created but I am still struggling with getting MJ to create an image with him back on to the camera and looking in to the distance. Any advice for the prompt?
Honestly, that shot is probably one of the toughest, I'm afraid. No simple solution 🫤
You did a great job explaining all of this. As an educator, I'm looking to see how I can use these tools to create more engaging content. Great job!
Thank you so much. Means a lot coming from an actual educator 🙂
Amazing application! So dope
Indeed
Christian when is your course coming out?
Limited alpha release is VERY soon (only 25 spots). And I mean VERY VERY soon.
Wider release for everyone will hopefully happen a few weeks later.
@@TokenizedAI let us know here please. Make a vid to announce it.
Are you on the mailing list? Make sure you are signed up for that. That's where I announce it first. I won't do a video on it until the full public release.
You can find the link to it in the description text of all my videos.
You can use the character's photos and a stick figure to get consistent characters
I know how it's done. Id didn't do it in this video for the reasons I stated.
@@TokenizedAI My statement was not true, but the results were hilarious. Would not be a bad video idea.
Is there a way to download the comic book?
Link in Description
As always, a great video Christian ! I have been working on something to your project and also wanted a certain "look" or "Style" for the characters and background . To get something similar to what I was looking for, I took an image with a look/style that was similar to what I envisioned, and used the " /describe " function in MidJourney to provide a prompt style description that I could use. I then adapted that style description for my needs. Sort of like reverse-engineering the style and look from something to fit my needs.
The results were kind of close but still needed a lot of re-rolling ! 😀
Re-rolling is underrated. I think a lot of people believe that it's a sign of bad prompts but it's not. We underestimate how much randomness plays a role in all of this.
Hi NICE VIDEO BUT hwo do i change to GPT4? i need to pay?
You need ChatGPT Plus
This is just fantastic, thank you so much for taking us through this and helping us.
Do you have a guide or video that shows us how to put a title onto an image
I recommend watching some UA-camrs who focus on image editing.
For all the people complaining, sit down. This is a tool for Artist, this will not end the Artist era that is completely impossible I am a Artist myself, this makes it much easier to define and make story telling way better for your comics, novels, anime.
Amen 🙌🏻
Nice joke, China game company already fired many game illustrators :) This is not a tool, its to slowly replace digital artists. Traditional artists are safe for now but nothing digital will ever be safe and 3D will be next. These kind of jobs will be hired even less along with design cause instead of needing 6 workers, now 1 worker can do all of them with the help of this new "tool". Writers themselves won't need to hire any illustrators cause AI exists
A big problem on the comics business is that there’s very few readers. To me that would be the attack angle, can we create AI readers? Any tips on getting AI to read comics? (Serious question)
Do you need readers or someone who buys the comic? Cause AI ain't gonna pay your bills 😅
@@TokenizedAI That’s the thing though. No readers no money. Money is digital, so why shouldn’t we develop AI that are consumers? What’s the reason not to? Sounds like a better idea than developing AI that makes stuff there’s no demand for. But money is always in demand.
That is exactly what i was looking for Thanks
A trick I use for creating consistent characters was to base them on real actors. It works best when it only looks70-80 percent like the actor so there's not a copyright problem.
How do you add their photos in mid journey
I can do character profiles and draw them. I can can do scenes and the overall plot.
Dialogue, fleshed out story, character development for everyone not that main character and why everyone gets along or not in a way where they don’t just feel like scripted npcs and all of that is beyond me.
That's quite a bit already though 🙂
@@TokenizedAI All of this time I’ve been caught up and stalling because I can’t do the heart of the story, the character interactions and dialogues, the journey as you gave as the premise in your example. You get that right, and you almost don’t even need the elements that I can do.
But, you just made me realize that a drama / opera isn’t the only way to tell a story. You can take a narrative approach. Like a journal entry or summarization of the days events or life story and keep the in-depth character details and lore separate.
11:55
If I hadn't already subscribed, this would've been the reason! 🤣 Great video. Thank you!
in ogni immagine la faccia dei personaggi è diversa!!! Come potete sostenere che potete realizzare qualcosa di serio con la deficienza artificiale?
Perché non sono un pessimista così incredibile come te che apparentemente ha bisogno di trovare qualcosa di negativo in ogni cosa. Perché non ti godi la vita così com'è? 🤷🏻♂️
Hello Mate... Quick Question... Can we have Consisten Buildings / Interiors?
How do you define a consistent building or interior? When do you consider it consistent and where's the point at which you longer consider it consistent? And why is the consistency important for your use case? Is consistency really that important for what you want to do?
Your prompt defines a bunch of inputs that the model then tries to approximate via a diffusion process. It's not a 3-dimensional blueprint of a building with furniture placed in certain locations and shots taken from different a angles.
I am so happy I came across your channel recently. Ive been up all night, ive subscribed and downloaded the links awesome stuff. Im about to open the zip file from comic book, I feel like Ive just been given something really special and totally free. i especially love the midjourney cheat sheet ive been looking for something like this. You deserve a Medal.... :)
Thank you so much.
Thumbs up once again. The best tips on comics and midjourney. The most valuable tip for me was that the characters don't have to be 100% coherent. I took a look at some of my old comic books and lo and behold, the cartoonists didn't exactly cover themselves in glory when it came to consistency back in the day (in the golden age of comics in the 50s). A blonde in a red dress and a brunette in a grey costume met a redhead in a grey suit and a black-haired man with a moustache in a blue mechanic's suit. So you have to differentiate the characters well, then it should work. I started my own project tonight (a Victorian steampunk comic)....
Please do share your progress on your project. I would really like to compile all of the great work spawned from these videos 🙂
Thanks for making this video.
A small idea for a video, the new permutations should be doable without pro, since it uses no new feature, I think a simple Discordbot should do the same trick.
Standard members have access too.
@@TokenizedAI strange, in the officehours I thought they said pro only, maybe they changed their mind
Yes, they said that initially but changed their mind.
How would you take this one step further and publish it as a physical copy? What sites are available? How to upload your panels?
Check out KDP.
@Tokenized AI by Christian Heidorn
KDP (Kindle Direct Publishing)
It looks like you can publish ebooks and paperbacks on there, thanks!
Pleasure
Hello, Christian. As usual, a polished and sleek video with all the necessary documentation. Thank you very much; I've been following you for quite some time and am continually learning new things from your videos.
That's so great to hear 🙂
THAT WAS AWESOME . THANKS
excellent work!
Thank you! Cheers!
What if I want to create consistent SCENES/ENVIRONMENTS rather than just characters? E.g. a character's bedroom that he returns to at different points in the story. How can I ensure I can get the bedroom to look the same or similar? Would I follow the same guide as consistent characters i.e. generate scene images and then paste in multiple URLs? Or is this impossible? Thank you
Two problems:
1. Everyone means something different when they say "consistent environment".
2. You can't really do that because MJ isn't actually aware of space. It doesn't know it's a room. So it can't show you the same room from different angles. It can show you a similar room from another angle, but not the same. If you use image references, you'll be able to transfer the style of the image and to a certain extent the interiors, but it will always be a different room.
@@TokenizedAI There is a way forward, but you'll need a 3d program like Blender or Daz Studio. Arrange a room scene, render it, then send the image to MJ. This way you can get the same room from within a cube space, although AI may make subtle changes to the 3d render depending on your prompt.
Great man! I think you know this already but IF we wanted more consistency (which in this case is NOT a problem as they look consistent enough) there is this old trick of using a celebrity BUT changing his/her gender to opposite or change his/her age or even style like anime style etc. That does not solve the different clothes problem 100% but maybe there is a solution for that too. Also using already created images as an image source could be helpful.
Many thanks! Looking forward for those files. Great video!
I’ve never used midjourney yet personally but I think you are right. If you used image reference prompts like for the character and background I think that would help with the consistency issue. Would be a lot more work but nothing compared to drawing everything by hand lol. I think midjourney as whole has a language and process that’s yet to be fully discovered. I truly believe you could make a full graphic novel using this but just needs the correct prompts and processes in the correct way to make it possible.
Great video. Out of curiosity, what program would put all of your pictures into comic book format?
This was extremely helpful, and I really appreciate that. You were willing to take the time to share all of this with people like me who love this kind of stuff and love this kind of use, comic book, usage of AI. Thank you, and will certainly be providing my email address for, these amazing prompts.
Just trying to create helpful content 🙂
Nice video Christian. Feedback: it seemed you were angry about the feedback you receive about the "consistent character" issue. Instead, make a joke of it or tease and have fun. Love your videos, thank you.
I wasn't angry. It also wasn't feedback that I received. I was challenging the people with unrealistic expectations, who expect Midjourney to be able to do literally everything under the sun, when only 5 months ago none of them were able to do any of this because the tech wasn't even available.
People are too impatient and I think it's more than fair to say so.
Amazing work-thanks for this! Do you think MidJourney and other text-to-image tech will eventually be able to make consistent characters an integrated feature? To be able to save a detailed description/look of a character and have the AI put that same person in whatever scene/style you choose seems like it should be in reach, where it can do it pretty decently with famous people. But I lack the understanding of how it works..maybe I’m just daydreaming :)
Yes, but not in the way you might think. O believe MJ wants to eventually allow people to create their own mini-models for consistency. But I think that's still far away.
Super Video, danke. War sehr aufschlussreich. Beim Dialog hat GPT-4 aber noch Verbesserungspotential: Der Dialog startet immer mit Maya und die Antwort ist meist etwas einsilbig. Das bringt nicht gerade ein spannendes Story-Feeling auf. Aber so als Baseline und Inspiration schon mal nicht schlecht. Vielleicht schafft es Midjourney in einer der nächsten Versionen sogar, konsistente Charaktere zu generieren (ohne, dass man Midjourney zuerst mit ihnen trainieren muss). Dann könnte man sich, statt fernzusehen, UA-cam zu schauen oder ein Comic-Heft zu lesen, ein eigenes Comic-Heft erstellen, von dem man die Geschichte noch gar nicht kennt. Sicher witzig. So ähnlich, wie sich selber eine Gute-Nacht-Geschichte erfinden/erzählen.
Toller Kanal. Weiter so.
Viele Ostergrüsse aus der Schweiz 🇨🇭
Ja, der Dialog ist etwas hölzern und Leo kommt etwas "dumm" rüber 😅 Allerdings muss ich für die Videos alles ein Stückweit knapper halten. Es geht ja primär darum, ein Proof of Concept zu liefern 🙂
thank you bro. that was really helpful
I was thinking of this idea the other day and now I see it on my yt feed..
You are a god in teaching online!!!! Wow I am so glad I found this channel
Wow, thank you!
But the characters looks different in every image… are they different people?
Are there anyway to make the same character in each scene?
Yes and no. It's complicated. Watch my character design series.
Some great prompts!!
Thanks 🙏🏻
Hey, I'm also trying to grasp the idea of having multiple characters in one setting but I've been at it for like 2-3h and I can't seem to get any results, I get only ONE character every time. The prompt I've used was this: /imagine prompt: Blend of comic book art and lineart in full natural colors. In a bustling medieval city, a wise, middle-aged paladin with a white beard and long, flowing hair who wears ornate silver-and-gold armor and a blue cloak and a young paladin in his early twenties with long golden hair who wears royal blue-and-gold armor with a matching cape meet each other at the city gates --v 5 --ar 9:4 --s 750 --v 5 --s 750 --v 5 --s 750.
Am I doing something wrong? I've been following your guide pretty much step by step.
Why are you adding the same parameters 3-4 times?
I really recommend you watch some of my videos on syntax first. I also have a video on multiple characters.
@@TokenizedAI sorry, I think it was just a copy paste issue (i’ve edited the post a few times to make it look solid but I didn’t notice I’ve added the parameters multiple times 😩
I’ve seen your multiple character video aswell, it was actually the video that inspired me to go on this journey. (I work as a surgeon but I’ve always wanted every since I was a kid to create my own comic books… so thank you for all the great info!)
Great video, my current project is actually taking3 short stories I've already written and trying to use MJ to create not just a better front and rear jacket but create a read along either comic or some other medium, obviously this is a lot to learn but your video is literally what I'm attempting right now, I am also following on Twitter so I will comment there at times also maybe with more Q's
Awesome. Make sure to share your progress on the project 🙂
Thank you for this amazing video!
My pleasure!
20:28 That is a dangerous looking snake.
This ks actually a great tutorial. Thx!
but you have to pay $20 a month for GPT-4 right?
Yes
Real informative great help.
Wonderful video, thanks a lot for the info.
Glad it was helpful!
Thanks for sharing these incredible tips!
Anything you use or “create “ using anything AI can’t be copyrighted .
What exactly is your point? Like...
seriously. When was the last time you actually enforced a copyright on your creation?
90% of what the average user creates gains absolutely nothing from having copyright. This obsession with copyright is so strange. Especially among people for whom it's essentially meaningless beyond personal preference. Unless you've created intellectual property that has widespread recognition and brand value, you couldn't care less about having copyright protection.
Here's another example: If you create a comic book and outsource the illustration to an artist, you also do not automatically hold copyright on those images. You get usage rights, unless of course you coerced the illustrator into a raw deal. In some countries you can't even transfer copyright at all. But you can transfer usage rights. And you do hold copyright on the overall comic book creation.
So please enlighten me, how is this copyright discussion relevant in 90% of the practical cases beyond it being a meme at this point? It's an academic discussion amongst lawyers at best.
@@TokenizedAI Im Not talking about legalities , in talking about self publishing companies cracking down on AI . They are not accepting anything that’s been AI generated . Don’t lie to these folks who think they’re gonna make thousands of dollars .
Спасибо тебе большое за видео! Я долго искал такую полную инструкцию.
What’s the name of the font used on the cover?
Amazing video. Fantastic channel. Thank you for opening my eyes and my mind. Quick question-how did you create that awesome title font for your comic?
I did that in Affinity Publisher. Check the supporting material for access to the files.
👍this is so cool thank you
I watched this video. Good prompts, great info. But I have a few suggestions, to create consistent characters, I'd create them in various poses on white background first, then remove the background and save as png files. Then I would create separate backgrounds of forest, river etc and place these character png files in the foreground and create dialogue boxes to have consistent characters throughout. It takes a bit of work, but it is worth it in the end. You'd have the same characters throughout making for an effective comic book.
I'm aware of how to do it and also believe I even said so in the video. I have an entire series on character design on this channel. This was just a demo and by no means meant to imply this is the whole process.
As you say yourself. It takes work. Hence why it's far too much to stuff into a YT video. People tap out if the video is much longer than 15 minutes.
Thank u for your content, it's amazing! Soooo 😂 you said "go learn how to draw..." And I actually did it many years ago, and I can draw, I'm a professional illustrator, and I'm watching your amazing tutorials and do some experiments in Midjorney but I don't understand how you create a comic book for one day? I spent a few weeks just to get illustrations I wanted and they still need to be fixed in Photoshop. Ofcourse if I had to draw them by hands that would take me a month.... But still! I feel like my process is way too slowly that it could be...
If you pay attention to the images in the video, you'll see that the characters aren't 100% consistent. This is the compromise that I made and that's why I completed it in a day. If wanted everything to be perfect and also a better story, then of course, it would have taken much longer.
Great video. Unfortunately US law states that intellectual property can be copyrighted only if it was the product of human creativity, and the USCO only acknowledges work authored by humans at present. Machines and generative AI algorithms, therefore, cannot be authors, and their outputs are not copyrightable. (Hence, if you don’t mind your work being in the public domain, go for it.)
It's still very early in the debate. Laws obviously aren't static, so I wouldn't be surprised to see this space evolve dramatically with some time.
How they swap the face ?