PLACE a Character in ACTION Scenes - Midjourney Character Design
ฝัง
- เผยแพร่เมื่อ 2 มิ.ย. 2024
- 📙 Midjourney COURSE mastersofmidjourney.com Beginners to Advanced
🚀 FREE Midjourney Cheat Sheet tokenizedhq.com/freebies/mj-c...
🔗 FREE Promptalot Extension promptalot.com/extension
🔗 FREE Supporting Material tokenizedhq.com/freebies/vide...
Folder: 2023-01-31 - Characters in Action
🌐 Check out the full blog post:
tokenizedhq.com/midjourney-ch...
-
📰 AI Newsletter 👉 Your Inbox tokenizedhq.com/newsletters/ai
🐤 Follow me on Twitter / chrisheidorn
💼 Follow me on LinkedIn / christianheidorn
📸 Follow me on Instagram / christianheidorn
💬 Join the Tokenized AI Discord tokenizedhq.com/invite/discord/
🎬 Character Design Series 🎬
Playlist: • Create CONSISTENT Char...
Part 1: Create Consistent Characters • Create CONSISTENT Char...
Part 2: Place Characters in Action Scenes • PLACE a Character in A...
Part 3: Apply a Consistent Character Style • Apply a CONSISTENT Cha...
Part 4: Create Multi-Character Scenes • Create MULTIPLE Charac...
Part 5: Create Facial Expression for Characters • Create FACIAL EXPRESSI...
Part 6: Infuse Your Characters with Themes • INFUSE Themes into Cha...
📺 Recommended Related Videos & Playlists 📺
Monetizing AI: • Make Money with Midjou...
Activate GOD MODE in MJ: • Become a GOD in Midjou...
The SECRET Language of MJ: • The Surprising TRUTH a...
How to Use --SEED in MJ: • How to Use the --SEED ...
Ho to Use NEGATIVE Prompts in MJ: • How to Use Negative Pr...
-
So you've created your own character in Midjourney.
But now you'd also like to place your character in action scenes.
While it's relatively easy to add a background to an image, changing the camera angles and displaying your character with its full body is a whole different story.
In this tutorial, I'll show you how to orchestrate different parts of your prompt in order to place your character poses and live-action scenes.
This video is Part 2 of a longer series about Midjourney character design. Part 3 isn't quite ready yet, but it will be released within the next couple of days.
-
⏰ Timestamps ⏰
00:00 Bringing Carla Caruso to Life
01:25 Simplifying & Fixing our Prompt
02:39 Positition Carla on a Rooftop
03:17 Zooming Out the Camera Angle
04:00 Getting Carla's Feet into the Shot
04:36 Changing the Camera Angle
05:53 Placing Carla on a Horse
06:58 Carla Exploring the Inside of a Temple
08:00 Carla Roaming the Streets of Neo Tokyo
08:40 Carla in an Army-themed Comic Book
09:32 Dynamic Action Shots
10:01 Adding more Characters to the Scene
10:50 Applying a Consistent Style to Scene
-
#midjourney #midjourneyai #midjourneyart #aiart - วิทยาศาสตร์และเทคโนโลยี
![แข่งแลกของ 1 วัน!! ของใครแพงสุดชนะ!! [Ver.2024]](http://i.ytimg.com/vi/joRT59ilPHg/mqdefault.jpg)
These tutorials are not common and you explain them very carefully and in a practical way.
Thank you 🤗
This is the most in depth guide I've found on this kind of thing. Thank you for showing every single step with every detail along the way so we can really follow the WHY instead of the how.
You're very welcome!🙂
I💙AI! So far I have understood three things:
1. AI is phenomenal!
2. ChatGPT & Midjourney rock! .
3. All these are not for dummies .They are for creative people, used to work with the medium.**
***You know - the ones who already work with CAD systems, MAIA,RHINO, Photoshop, FinalCutyPro, etc and know a bit about Javascript and Python too, among other things. Even though you don't need all of these here - you have to be able to plan exactly ahead of time what you wantto do - that we call DESIGN for short. Otherwise - you are on the mercy of the generic images hoping to be unique....That will never happen.
Love your tutorials - intelligently spelled for sharp beginners. Thank you!
You are a good teacher. You go through each point step by step. Showing the prompts for each modification is good.
Thanks so much for your efforts
Glad it was helpful! 🙂
yes thanks, very easy to follow along.. but question... are u ising v3? i cant use weights in v4
I've been using MJ for film production look books / pitch decks. I"ve found that putting consistent character together into compositions , multiple consistent looking characters in one composition, is the advanced-level challenge I'm currently trying to solve. I have to scoff at the people who think using these AI's are extremely simple, easy as typing in what you want and getting it. Far from it. I'm using MJ, DALLE2 and Photoshop to mix multiple images into one composition, for most of the images we end up using for the book. There's way more work involved than most casual observers would imagine.
Yep, that's why I'm also starting to refuse to engage with people who are just upset because they feel like their profession is at risk. All they see is that any average dude just needs to enter 2-3 words and MJ will output a surprisingly good-looking image. What they just don't understand is that as soon as you have a very specfic idea in mind, it's not that easy anymore. And I'd argue that most of them wouldn't be able to figure it out. It's a tool like any other that requires skill and that can only be learned by using it extensively.
I know what you mean. I get comments on my videos saying "the Ai did a great job capturing so and so's likeness" ... Excuse me! I tweaked the prompt and rerolled dozens of times gradually honing in on the right look
Playground or Leonardo AI has a scene generator tool I believe where you can merge scenes. It’s pretty dope. You could take your MJ pics and put them in there to see what it does.
I solved this. Dm me if you still need it.
I have to say that your videos are really awesome. I rarely see this teaching quality and attention to details as you do! Very helpful
Thank you! Really appreciate the kind feedback :)
This man is a king.
🙈☺️
Amazing ,thank you sir, cant' wait for part 3 .....the best tutorials on youtube by far, your help is much appreciated, thanks again .
Much appreciated! 🤗
Fantastic job! And you are hitting on exactly what I want to know. Can't for the next video, thanks!
Awesome, thank you!
What wonderful content! Thank you
Thanks so much! 🤗
You are a God!!! Thank you so much to share all your knowledge on this subject!
I wouldn't go that far, but thanks anyway 😅
Love your content! This is a huge help.
That's great to hear! 😊
Brilliant! Keep it coming!
Thanks! Will do!
thank you a lot for the very detailed guideline, hope you will provide many helpful videos like this in the future
Working on it.
Wonderful tutorial ! Thank you so much !
Thank you. Very interesting techniques!
And there are so many more! :)
This was so great! It was really helpful to watch how you work thru editing the prompts to get the results you want. These tips were extremely helpful! Thank you!
It's those small changes and the differences they make that really help people learn.
That's why I personally dislike videos that just show you a bunch of nice (and most of all, curated) images, which usually implies that it's very simple to do. When it fact, it usually isn't 😉
@@TokenizedAI YES! The thought process and intention, knowing what to change and tweaking things when it's aaaaaalmost there! Learning how to navigate those things is invaluable.
Thanks for putting yourself out there to share what you know and experiment on what you suspect. I signed up for interest in your mid journey course. Been following along with the examples you share in your video and I’ve learned so much in the last Couple of days when I started.
I really need to stop procrastinating and start recording the course 🙈 I hope the free content I'm putting out makes up for the massive delay.
And thank you so much for appreciating what I'm trying to do 🤗
Great video😻 love this style which is very easy to follow and clear. Thank you for making these videos 🥰🙏
That's really great feedback, thanks! 🙂
This is super cool, thanks a lot for sharing those tips
No problem!😊
Such great videos Christian! I had to tag one of your videos to mine👍
Thanks a lot! :)
Great stuff, thanks a lot! 😍 Looking forward to try this on my own projects!
If you're working on something bigger, don't forget to share your results once I get our little community up and running :)
This is incredibly helpful, thank you so much! It's just mindblowing what can be done with these techniques
Glad you enjoyed it!
This is fantastic expert level content man. Keep it up, subscribed!
You are making this so easy and awesome. thanks!
You are so welcome!
Hi christian, nice followup. Thx for sharing it's really an interesting subject you treat in this series
Thanks! :)
thanks for your content man !
I had another way of doing quite the same but with a completely different technique :
I prompted my character alone in a white background in portrait, medium, full body.
and used it within the image prompt, it give a " as consistent" result aswell ;)
Yep, been experimenting with that as well. Looks like that works great too.
@@TokenizedAI it may give different intonation to the result, but you can always modify the picture used for the scenery prompt.
thank you so much for this!
You're so welcome!
the potential is insane
Awesome work!! I was stuck on creating actions scenes for my characters, your work is truly amazing, keep it up and don’t listen to the trolls out there
Thanks a ton! Really appreciate it, especially that last bit 🤗
Finally someone talks about this
great video, keep up the good work
Thanks 🙏🏻
Amazing !
You still have to play around to keep the character. I am sure, very soon, that the AI interface will integrate some options to get consistent styles and characters through further images.
AI is moving so fast !
Yeah, all of this is far from perfect and requires a lot of additional work. I'm just trying to figure out where MJ's current limits are. I still think a lot more can be done now already, . You just might have to use a multi-step approach.
Excellent video
Thank you very much!
Great Tips Christian 👍
Thanks 😊
Glad I was the first to comment. Love your content
Legend! 😁
My level has changed well and that's due to your guidance, I have some problems like clothes change and some scenes don't appear, and sometimes the style changes, but we'll watch the rest of the series and see where we go, thanks great effort
I struggle with those too. Don't forget that we're still doing all this in "hard mode". Wait until new version of MJ come out. All of this is going rock our worlds.
@@TokenizedAI Really!, excited about what's coming but I found in your other clips some solutions, especially that I'm preparing my own comics.
Nice. Thanks
Most welcome 🙂
HURRY and sign up early for Professor Heidorn’s class. It always fills up fast at Kick A$$ University! Thank you sir for sharing.
😅
thanks you!
You're welcome! 🙂
Wow, what a nice easy to understand serie. Fascinating stuff.
Little feedback, no need to re read the recurring boring part of the prompt :)
Too bad the AI has trouble fitting items into hands, like the torch. Maybe it needs an additionnal prompt like "a single torch held in one hand only".
Thanks! Yeah, hands still are a major problem. Only way to really fix this (in the short term), is by fixing them via inpainting either in DALL-E or Stable Diffusion. Though I believe DALL-E may be a bit more effective, given that MJ v4 actually uses Stable Diffusion.
"as you can see this is still the same Carla Caruso" , meanwhile I'm sitting here like "bruh, the face is TOTALLY different" :D But I'm not here to nitpick, this is still quite impressive considering that this is just the beginning of what generative AI has to offer^^
Yeah, fair enough. I was mostly referring to the overall character. You can fix the facials later on by adding image prompts. Purely text-based it's not that easy 😁
@@TokenizedAI not throwing shade at you, you did great :) Gonna be really interesting tho once you can do more specific things than just a generic looking character , e.g. adding tattoos or intricate design patterns of clothes.
@@thomasmann4536 Totally 😃
You are absolutely my AI mentor❤❤❤
What can I say. I'm humbled 😊
Thanks to make this :). Midjourney is marvelous and very creative BUT it's really hard and almost impossible to keep a real homogeneity in character or scene (as a watch your video, its still a problem). But if you don't expect something very accurate, thats a fantastic AI. Especially for the creativity. I can't stop watching all the beautiful things people produce... (I've tried for hours to produce a Twi'lek of Star wars and i can't succeed...).
Yeah, it's still not perfect. However, I'm 100% that it will continue to get better over time and we'll eventually be quite surprised how much we'll be able to control.
For now, we beed to work with what we have. 🙂
@@TokenizedAI I totally agree. Have a nice day.
It is so cool what these small little changes will do to the pictures generated. The tip with more people/living creatures in the panels is also quite cool! If you replace the seed through rerolling can you then later/another day still use a seed from before or is it then totally replaced and gone?
Seeds are just an input variable. So yeah, you can re-use them whenever you want.
Great tutorial again. Thanks. If i can give you a tip too -
use " CANDID image/illustration" somewhere in the prompt, this ill give you different body posses. Cheers
Interesting. Do you know why? 🤔
@@TokenizedAI I have no idea, I saw a "colleague" to
using it and decided to try it - and it worked :)
thanks
Love your vids! Do you offer tuition over zoom?
Thanks! Actually haven't thought about that before. Probably depends on the scope and what your objectives are. Send me an email at chris [at] tokenizedhq [dot] com and we can discuss.
It doesn’t make sense to put her name in even the first prompt if she’s not a known character. Furthermore, I don’t think adding the seed to completely different prompts (changed situations, environments, camera angles, etc.) makes any difference. As I understand it, the seed only is the value used to generate the initial noise pattern. If you use the same seed for the same prompt you should get very similar results, but change the prompt and you might as well forget the seed value.
Also, I did an experiment recently and I think the seed parameter may be getting ignored in V4. I upscaled an image, got the seed and then used it with another prompt, but when I checked the seed values for the resulting images they had changed.
You might believe that, but feel free to do some very structured tests and you'll see that it does make a slight difference. It's obviously not the same as using a known character because you can actually recognize a known character. But the unknown names end up being a blend of many different famous people who either shared that first or last name. And obviously, that makes sense since it's a scrape of the internet and who's photos show up most? Right.... celebrities.
With regards to your statement about seeds, it's factually incorrect. v4 absolutely supports seeds. The reason why your results are confusing you is because upscales use entirely different seeds than a set of 4 variations. Which is logical, isn't it? How can a set of 4 images from an /imagine command share the same seed as a single upscale? They can't. So you're comparing apples with oranges here.
Thank you so much for these incredible walk-throughs! Can't thank you enough for helping us visualize this in a way that no one else is doing. Explanations are so incredibly thoughtful and helpful. My question is, is it possible to create a character off of a real person? My attempts keep getting me variations that don't look like the original image at all. Thank you again have an awesome day!
I've fed Midjourney a few prompts including 'confused Norm MacDonald' and it handles those very well 😂 I imagine using someone for whom the bot can pull a lot of results makes a big difference. As far as feeding it images in the prompt, have you tried using images of the subject from different angles? Hope this helps!
It's generally very difficult to do this off an external image. It's much easier if your core character idea originates from within a MJ prompt.
@@TokenizedAI Ok thats great to know, I'll get to work on the trial and error process. Looking forward to more of your content to help guide me. Have a great day!
@@BilboShoze I have not tried that, but thats a great tip and something I'll give a shot today! Cheers!
This is so cool ! I do have a question though, can I use BlueWillow in this process? It is the tool I am currently using and as a beginner I need all the help I can get. Please share a tutorial on it in your next video. You would be a lifesaver!
I don't know because I haven't used BlueWillow much.
Great video im just having trouble on keeping the same clothes, than i think i might have a good handle on it.
You might find the next video useful then 🙂
@@TokenizedAI That's what I'm waiting for now ahahah been trying to figure it out on my own, thank you.
Well, I don't know whether it'll directly solve your issue but I'm pretty sure it'll help you figure it out somehow.
This is incredible ! 😃Does the process take long to generate? I am using BlueWillow and I wonder, can I use this tool to create a character design? Please share a tutorial, as a beginner with this new tool, I would really appreciate your help.
I haven't used BlueWillow much but from what I've seen, it can't really keep up.
What's really crazy to me is the fact that, even though her facial and sometimes body features are not exactly the same from scene to scene, my brain still recognizes her as Carla either way and not as a different person simply because her hair is close enough and her clothes are similar enough.
Psychologically, what might this say about culture where there is a precedent set on a group of people wearing the same types of things and having roughly the same between each other being perceived by outsiders as all effectively being the same person and being treated the same.
I actually think you're on to something here. Just think about how difficult it is for people who have never been to other regions of the world to actually discern the differences in people's faces. And I would assume it's the same vice versa.
@@TokenizedAI Woah, I hadn't even thought in that direction but I would not be surprised that's exactly right.
@@TokenizedAI you know, I wonder if there is a way to experimentally test this with Midjourney in some way. It's a bit late in the am for me, but I'm going to sit on this thought and see what sticks.
just amazing! Do you think the same is possible with Stable Diffusion?
Sure, why not? I just don't know what the right prompts would be 😬
I'm trying to make a Marvel fan film using AI and this video is very helpful. Although I'm using Playground AI instead of Midjourney. I'm not sure if all the tips apply but the general knowledge is helpful and I'm looking forward to the next video.
I haven't tried Playground AI yet but I suspect some things will work in a similar fashion. Do you know wether Playground AI uses a natural language processor on top of Stable Diffusion or is the prompt input direct?
@@TokenizedAI I don't know, I'm sorry. I'm relatively new to AI.
@@TokenizedAI Stable Diffusion understands the prompt's semantics, and there are operators to group and isolate concepts from each other and prevent "cross-bleed" between concepts, such as in this video how you made sure gray color only appears on the jacket.
This brings up the question why did you not use this prompt:
Carla Caruso, ...., bright red hair::1 Carla wears a gray leather jacket::1. ...other stuff...
Fair question. The main problem with that is that it tries to blend all the concepts together, sometimes losing important stuff unless you keep repeating the same elements in each segment.
I got much better and more consistent results with a single segment. Multiprompts are great, but they're not always the right tool for every occasion.
Thank you very much Chris. is there any chance i could learn about part 3 a little bit earlier than "a couple of days"? LOL, i am currently working on a comic book which is due on this Sunday, if only i could learn how to place a second protagonist in the picture and make him interact with my heroine while maintaining both characters' traits and detail, that would be great……wonder if i have the luck to learn it from you personally(maybe through messages?）. Anyway, your video have been of great help, appreiciate all your efforts, keep it going!
Part 3 won't cover that yet, since it's really tricky and I still need to figure out how to do it reliably 😬 Controlling the look of 2 protagonsists is really hard. Part 3 is mainly going to cover maintaining a consistent style once you've created all your scenes.
@@TokenizedAI Love the content. Like 唐华汉, I'm eager to see Part 3, but I don't have a Sunday deadline. Does the method you're planning to cover have a name and is it documented in any of the MJ documentation? Would like to read ahead...
@@TokenizedAI ah i c, thank you all the same! looking forward to this series, will continue check out for new information. BTW, you said there is a group where we can discuss such questions and together making progress, how can i join that group or club? would love to share as well
@@vegandilemma1129 Most of the stuff I show here isn't documented in the strictest sense. All of the command are obviously documented, but how to put them together and in which sequence isn't. There's no "template" that just automatically works in every situation.
@@user-jm8lb9lx8u I haven't set up the community Discord yet. But it's on my todo list.
Can I upscale and make variations of single pictures in that process? Or would that ruin the seed?
There is no direct link between seed used for the image grids you get from the /imagine command and the seed used for an upscale. So that probably wouldn't get you where you want to go.
To achieve a wider angle, I also like to use DALLE ai Image extender after creating an MJ image
You mean Ouptainting?
@@TokenizedAI it’s ‘Add generation frame’ in DALLE
Actually, the process is called "Outpainting". "Add generation frame" is just what the button says in the editor. "Inpainting" is the process of "erasing and replacing" a part of the existing image with something new.
Is it possible to create a consistent environment? Thank you!
I think it's very tough. Many have asked this though.
This is great... Can you explain the "seed" and the dashes --- please?
Please check my video on what --seed does.
I still don't get the whole seed thing. When I use any seed and comment with the envelope on the generated image, it showes me a clompletely different seed every time. It doesn't matter if I use reroll or not. The seed changes always. Any solution?
The seed always changes, unless you've set it to a fixed number. If you're not setting the seed with the "--seed ANYNUMBER", then it's obviously going to change every time. Also bear in mind that seeds will be different for a set of 4 images generated with the /imagine command than the seed that you get from a single upscaled image. They are different.
@@TokenizedAI What is the difference between using a fixed seed? Can I set a seed right at the first prompt? Do you have a video or can you address this subject better? What are the advantages and disadvantages of using a fixed seed? I would be grateful if you could respond to my comment and congratulations again for your great classes.
Teacher of the future making art of the future 🔮
🤗
Great Video, makes fun to watch and I learned much. Maybe I missed it but why did you train her name "Carla Caruso" and left it out in this video?
The more you zoom out, the less it matters. Also, to be fair, using a name isn't strictly "necessary" to begin with.
@@TokenizedAI Thank you, it's great to get your answers that quick. I love YT because of people like you!
I'm a bit confused, why don't you use text weights in these prompts while you talked about the importance of it in a previous video?
Ha, good catch! Text weights are relevant when you can't seem to get the type of composition you wanted. In this case it was working quite well. I'll cover multiprompts and their applications in some other videos though.
Is it possible to get two or more characters interacting in a scene or is it necessary to cut and paste them into a background scene?
You can get multiple characters to interact with eachother, however, it's super difficult to do if you have multiple protagonists. If you have 1 protagonist and 1-2 generic characters, then it can work.
@@TokenizedAI I want to illustrate a book. I’ve got a man (protagonist) two children and a mouse. I think I’ll have to think carefully how to frame each scene. So I’d need to do your seed method for each character? Can I have multiple seeds in one prompt?
From what I've heard, this method is less effective with non-human characters.
You're going to have a difficult time getting multiple protagonists that have a very specific look into the same image. I honestly haven't tested it much either. I might be able to provide more insights in a week or two, but now right now.
Alternatively, depending on how your illustrations look, you might be able to screate an empty scene first and then create the characters and their poses individually. And then photoshop them into the scene.
Not ideal, I know. But that's all I can offer right now.
thank you for your share, how to make same color for character clothes
MJ is still bad at assigning colors to specific clothes.
I was created exactly by your recommendations to promote book about AI replacing humans, which was written by AI.
👍
wow many thanks , im learning a lof with u, could do a prompt tuorial for PlaygroundAI , im poor and i cant pay for midjourney
I'm currently focusing on Midjourney but I'm sure I'll get around to Stable Diffusion or Playground AI at some point.
Awesome!!
Thank you! Cheers!
Why your pictures looks so clean all the time. My images very often look strange and like something went wrong
What do you mean by "clean" and "like something went wrong"?
@@TokenizedAI for example faces, especially with animated characters look not symmetrical and bring some distortion, but i make same like you ?
So I'm having trouble... I have my character but had an image link in the original prompt. Now I'm afraid it's too strong to allow me to change the environment. My character is pretty well defined with the right clothing, but I can't change the environment or pose even with weights
Sounds like either your image prompt or your character description is too dominant (I'm assuming it's just 1). Have you tried creating a second image reference that provides more context in terms of composition? You should probably watch part 3 of this series too.
@@TokenizedAI I've tried dropping the image links but then I lose face... I could rebuild him without the links but that'd be a shame! I'll give it a try, as is the links do seem to hold MUCH more weight than prompts
Unfortunately v4 doesn't allow us to set a image weight anymore. Otherwise you could have adjusted that. Maybe that'll come at some point.
Can I set her in profile view and back view?
Yep, that's possible. You'll just have to change the original prompt quite a bit. For a back view, you'll also have to avoid making references to her face (incl. "beautiful") because as soon as you do that, the AI thinks that it's an important part of the image.
But is there a good method of putting your character in a FIGHT scene? I've attempted two guys duking it out, or a martial arts theme ... MidJourney seems to place the characters farenough that it doesn't look violent, which defeats the purpose.
MJ isn't very good at fight scenes.
Does the seed somehow tell Midjourney exactly what the character's face looks like?
No, not really. The seed is just a value that adds "noise" to the command. When you don't set a seed, it uses a random value. If you set a seed, it keeps the "noise" value constant. While this might lead to consistent facial features, it doesn't strictly determine what the face looks like. That also depends on your prompt.
It would be nice the midJourney would follow the story , follow the same character ( face , body )
Hello does this work on using a photo on the promps?
Can you define what you mean by "does this work on using a photo". That's a bit vague and I'm not really sure what you have in mind.
@@TokenizedAI because I'm planning to make a comics made out of the employee's faces I will make them a kinda like a superhero and I'm using their photo I already have portraits of them and I wanna make them pose on different scenes for the comics
Ok, so if I understand your original question correctly, you want to use image prompts to place your employees' faces on the characters in the image?
That really only works if you're doing it for 1 person at a time and you'll need multiple images of them from different angles. Are you sure your employees are OK with you feeding MJ with their photos? 😬
@@TokenizedAI yeah they wanna use MJ for marketing purposes planning to release a comic book and change its marketing approach using MJ cool stuff
@@raphaelamido8785 Doing it for 2 or more protagonists in the same scene is difficult (maybe even impossible right now).
As I said, you'll have to feed 3-4 images from different angles as image prompts. Ideally also full body photos, otherwise it thinks that you want a portrait or close angle shot.
I would recommend creating separate images using a seed and then post-process them together manually in Photoshop.
Just the only idea I can think of right now.
I thought MidJourney didn't look at all like it was trained on Rob Liefeld's artwork... until he tried to get it to draw feet.
😂😂😂
@@TokenizedAI does this work for v5??
What is the seed # doing?
I explain it
Your end video link to "the entire playlist" doesn't seem to actually link to a playlist. :)
lol what? I'm looking at my screen in TH-cam Studio and it clearly says it's linking to the playlist, but then when I watch it.... you're right, it doesn't. wtf? 😅
Ok, I found out why. The playlist currently only contains 2 videos. The first video is already recommend on the left side of the screen. And the second video is the same video that you're watching. TH-cam won't display videos more than once on the same end screen. So that's why. Need to wait until I put out Part 3. Then it should work 😂
@@TokenizedAI Ahhh, mystery solved! I've enjoyed these videos, looking forward to more. :)
Great vid. sure wish midjourney knew how to draw hands though
v5 is already much better at it, tbh
I think the AI interpreted your expression of “full body” is not muscular or slender, but a little bit plump… You can see that Carla is shown here with the big hips section big buttocks in big chest and that is the word full body actually the word means you meant full body shot but do you understand what I’m saying?
I know what you mean but that's not the reason. It's due to the comic style and v4.
Now if there was a way to add multiple seeded characters to the same scene
Adding multiple "seeded" character is effectively impossible and I don't really see it happening anytime soon. At least not by using a seed.
How to include additional character?
That's more difficult. I'm trying to figure it out.
PART THREEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
Working on it 😅
now matter what i do she is not looking down; can repeat "looking down" or "facing the street below" or something like that but it's not happening for me......
Hard to say why without knowing the rest of your prompt. Other elements in your prompt might be too dominant.
so the key is iterative approach in building up the desired prompt
Always
Why don't we get more specific all right? Here is an interesting example. I have been trying to get a image about "one person sneaks a bite on a girl's icecream in her hand while she is not looking" but failed at every attempt. I bet you one dollar to realize it on your next video, haha. Of coz, cannot use img2img, must purely use tag and mj only, no ps as well
Sounds like a pretty difficult scene for the limited control we currently have 😅
@@TokenizedAI It cost me several hours with practically no progress at all, so I brought it to the man whom I think ranks top in this area, which means, you.😁
@@user-jm8lb9lx8u And yet, I'm still only a mere mortal. 😂 I'm going to give this a try once I work on the multi-character video.
No matter what I type, the character only is showed from a a headshot perspective
Difficult to say what's wrong without seeing your exact prompt.
Still big problems in V5 when creating hands...
Still better than v4 though
👍
🙏🏻
Not bad the eyes don't look good in full body shot comic book style.
wuoa! --ar in v4?
Did you think --ar doesn't work in v4? It's just limited to 1:1, 2:3 and 3:2.
@@TokenizedAI i know that ar doesnt work in v4
And yet, I did it throughout the entire video 😉 What makes you think that it doesn't work? Even the official MJ documentation explicitly states that it works. 🤷🏻♂️
@@TokenizedAI lol...Sorry... I try with 16:9, hihihi. Love your videos. Thanks
@@TokenizedAI only a preset selection of --ar parameters work, is what he's saying. You're both correct :)
After two days trial I start to think maybe there is no way trying to get 2 character interact with each other the way you want it. With poor prompt techniques the successful rate is merely 3%-5%. With improved skills maybe the rate can go up to 10%, but no more than that, at least not under v4 of midjourney. Pretty disappointing.
I do think it's possible, but I'm also pretty sure that you need to jump through many hoops before you actually get to your desired output. I'm testing all sorts of approaches right now as I type this. I don't think I'll have an answer for you before your due date, but hopefully I'll have a video ready early next week.
@@TokenizedAI Never expect a video this much before. Due day is not that important comparing to the importance of the problem-solving itself, and thanks again for your effort, much appreiciated
To be honest, it's proving to be really challenging 😔
@@TokenizedAI Yes...That's why I felt a little frustrated the day before. But I totally agree with you: things will definitely get much better giving it time. Just not right now.😮💨
Won’t surprise me if you are actually an AI teacher, 👏👏
🤣🤖👾
its never a consistent character though 😅
Depends on how picky you are. We're just trying to figure out what's possible within the limits of MJ. Part 3 will help in getting the style more consistent. But that comes with other disadvantages.
As long as something changes whenever you generate an image, it's pretty useless.
I guess it depends on what you're trying to achieve. Also depends on what you mean by "something".
I disagree with the statement that it's useless. It might be useless in your particular use case. In others, i's good enough. I'm just demonstrating what can be done with MJ. The limitations of MJ aren't my fault. If you have a better way of doing this, feel free to share with all of us. We'd love to see how it's done 🙂
Anyway, if by "something" you mean her outfit, then Part 3 might be interesting to you. However, I'm also pretty sure you'll find something else that doesn't work there.
This is new tech. If you expected this to be as easy as pressing a button, then I'm afraid this might not be for you.
Many thanks for your work, you have the BEST and most helpful videos out there. Thumbs up!
Can I offer you something free from my services! I feel obliged after watching your videos 😀
Thanks so much! 😊
Buddy, sorry you're making your self a fool... that isn't how seed work, it's totally pointless what you did with "emojis reactions" and re-running the seed... you're not "training" Midjourney in anything... please read the paper and the details of how it works... specially v4. Don't fool your self thinking a machine learning algorithm trained with several billions of images that your 15 emojis reactions does any difference to it... Try this: Remove the seed completely, and re-run all the prompts... you'll see the same thing... It's not the "seed" doing anything special... as Midjourney clearly says in their docs... The seed is JUST THE NOISE added to the source noise!!! So you're not TRAINING Midjourney in anything... and the fact that you can drop the NAME of the character is because it's a useless part of your prompt!! It doesn't match anything that Midjourney has in their data set, so as a useless part of the prompt it was ALWAYS ignored anyways!
Thanks for your comment. 🙂
Not really sure why you're referring to "training" right here because I don't mention it in this specific video. Literally not a single time. Everything you've mentioned has already been discussed ad nauseam in the comments of the video you are referring to as well as this one. I've also pinned one specifically for that purpose.
I am also aware of the fact that a seed is just noise. I don't recall making any statements specifically linking the seed to "training". Yes, I used the word "train" in the first video, though I specifically put it in quotation marks.
However, your statement that setting a specific seed has no purpose is factually incorrect. Since you referred to the Midjoureny Docs, please have another read through the very first paragraph. It contains this sentence: "Using the same seed number and prompt will produce similar ending images". So it does have a purpose and in the context of these videos, I don't see why you have an issue with that.
Anyway, appreciate your comment, even though it seems that I may have upset you quite a bit. Sorry about that. I hope your day isn't ruined because of me 😉
I'll be moving on now.
Someone sold before the pump.
For the original character, what style did you use to get the non-realistic "drawing"?
I'm pretty sure I share all the prompts in thr videos.