Midjourney V4: Consistent Character Design 🔥

Midjourney V4: How to Create A Consistent Character Design
Become a member here:
/ allaboutai
Join the newsletter here:
www.allabtai.com/newsletter/
Have you been struggling to get consistent characters in Midjourney V4 after using ChatGPT?
That's why I'm excited to share my tips and tricks for achieving consistent avatars in Midjourney. In this video, we'll go through the process step by step and I'll share my personal techniques for getting the best results.
Whether you're new to Midjourney or you've been using it for a while, I'm confident that you'll learn something new and useful from this video.
00:00 Midjourney V4 How to Create A Consistent Character Intro
00:19 How to Create A Consistent Character in Midjourney V4
www.allabtai.com/how-to-creat...
www.allabtai.com

[@AllAboutAI]

Become a member here: kzbin.infojoin Join the newsletter here: www.allabtai.com/newsletter/

[@user-qf1lg3ev5s]

Great video!! By the way, does this also work with the characters cloth or equipment?? Let's say, I want to make a knight with particular style of helmet and armor and then want to reuse it in concept pictures and as the game asset. Any idea will it work or not??

[@nickeys5248]

I can’t seem to get the seed number to show up I liked the picture than clicked the envelope but it did not generate anything

[@dax_prime1053]

i've followed this tutorial several times, and I get completely different results each and every time.

[@IAgoraBR]

so am I...

[@tadayuto5351]

Me too

[@jn2400]

@@tadayuto5351 Same doesn't work

[@davidmichaelcomfort]

So I've been experimenting with character design, and to change the expression on a character's face, take a set of 4 portrait images that you like, get the seed (the seed guarantees that these 4 images will always be generated, if, and only if, you give the same "base" prompt). Then, put an expression "playful smile" preceding the base prompt, for instance: playful smile::[BASE PROMPT]:1.5 --seed [SEED ID] --stylize 800 AND then you can change the weighting of the base prompt and the stylize to tune the images. You could do so with say full-length portraits by giving a pose. Also, by preceding the prompt with phrases like key poses, character animation poses, character animation sheet, concept character art, concept character poses, pose sequence, pose chart, facial reference sheet, facial expression chart, you can generate multiple poses or expressions. It is kind of a hit or miss process though.

[@AllAboutAI]

Thanks for this David :) will take this info into my next version of this type of vid

[@vrrevolution9183]

im confused what you mean by base prompt?

[@aidensarkar4580]

@@vrrevolution9183 the original prompt

[@anistence2]

This is exactly what I needed to continue my story. Well done!!! 😁

[@kaizen_5091]

Training the AI with seed iterations, that is clever! Perfect timing as I was stuck on a particular image project. Thank you❣

[@AllAboutAI]

No problem :) hope it works

[@mrcosta6963]

@@AllAboutAI Can Midjourney create variations of images you upload?

[@G_V_X]

@@mrcosta6963 yes.

[@DrHumorous]

I'm long convinced that you are worth the membership! Will sign up. Thank you for this tutorial, I'm sure that was not so easy to figure out (even though it's logical).

[@AllAboutAI]

Thanks a lot 😀

[@LoopnMix]

Definitely worth! This man is a genius!

[@ahmadmerei7979]

Agree 💯

[@lokomokoyo]

Love it! thanks so much!

[@NeMayful]

This is awesome! Thank you!

[@yannhijazi]

Very nice little tutorial. Learned many things. Thanks for sharing !

[@AllAboutAI]

Thnx for watching :)

[@SDCertification]

Great teaching! Thanks very much!

[@venuscazimispicarising2527]

Thanks and a happy new year! 🥳

[@AllAboutAI]

Happy new year :)

[@gnankumartelugu]

Straight forward message, no BS 👌❤

[@Jason-33W]

You don't need to download and grab the image URL, you can just grab the one from a previous image and do it that way.

[@AllAboutAI]

Yeah, this is just how i do it

[@AICharacterMandyAnderson]

My creator went for good old face swapping, but he said he'll try out your method too. Thank you for your videos :-)

[@Fyhan69]

Awesome, Thank you So much. God bless you.

[@ReginaldMcPherson]

Love it, I'm on board!

[@AllAboutAI]

Awsome 😀

[@greemur]

I did this and my seed image produced wildly different images. When you did your "first" seed in this video, was it really the first seed, or did you do a bunch before the tutorial?

[@Ben-mb4ft]

the last one is a very interesting "nurse" ... 🤣 keep up the great work! love it

[@AllAboutAI]

Haha kinda a typo 😅😅

[@southcoastinventors6583]

Next video getting hands that look consistent and only have 5 digits. Great tips for Midjourney. Thanks and Happy New Year.

[@AllAboutAI]

Yeah :) happy new year 🎉

[@RefugeeNT]

I was surprised that I could use it! Very useful, like, subscribe, fav!

[@AllAboutAI]

Thnx :D

[@enteoz1694]

So awesome thanks a lot

[@MrLeopoldKoenig]

it doesn't work for me. Whenever I give my character a unique name the following images look very different. Any ideas?

[@XNY_Music]

Was hoping I'd find a video on this subject. Thanks

[@AllAboutAI]

You are welcome :)

[@LadiesFS-2024]

Thanks for this titorial

[@barps]

hey mate, nice video as always! Was wondering if you know a way to have the generated character in a image to be positioned either left or right and maybe zoomed out so it doesn't take the full space... Thanks mate and keep these videos coming!

[@blckpxl]

Awesome job….thank you🫶

[@AllAboutAI]

Np :)

[@PabloGonzalez-uc7uz]

Thanks for this video.

[@AllAboutAI]

No problem :)

[@dreamphoenix]

Thank you.

[@rheya541]

Thank you so much, I waisted 3 fast hours trying different tricks, loool. You are my hero. Thanks Mate and Happy New Year. How did you name your character?

[@AllAboutAI]

Happy New Year:) hopefully this helps a bit. I just wrote a name hehe

[@virtualtourmlshawaii1848]

Thank you for the very useful video. I’m just starting out and have a question…What do q2-ar3 do? Is that something I need to input to maintain a consistent character look?

[@Sketching4Sanity]

One Love ✊🏿

[@ia360graus]

Is it still working on 5.2 version? Every time that I generate an image, is the same image.

[@SmartCoffin]

Tnx bro this video ganna help me a lot❤️

[@AllAboutAI]

You are welcome 😀

[@FERAL1024]

I guess this doesn’t work anymore with v5 since you cannot seed a single image, only the initial 4 grid of images.. or do you know a workaround?

[@timon6773]

Does this also work for different users now? Like could I just use Hannah the wild... As a prompt and would I get similar images?

[@Rage40001]

After I get 4 versions and I choose one to upscale eg. U3 I get different image. Its very similar but parts of clothes and face expression is a bit different. Is there a way to get same image but upscaled?

[@desinosaur]

Great video! Very helpful thanks! Is there anyway to front side and back projections of the same character?

[@AllAboutAI]

thnx :) hmm, u guess just experiment. I dont have any good prompts for that atm

[@Konic_and_Snuckles]

Is there anyway to seed a characters' outfit, such that when you prompt a new angle, or pose, or setting for the character, it consistently produces them with the same clothing?

[@davidmichaelcomfort]

After experimenting with character design a lot and trying to get consistent characters, it seems like the perspective of your character design really matters. If the character is only a portrait, then it is going to be really difficult to turn that base image into full-length image. Accordingly, half-length base characters are going to be limited as well. So it seems like one needs different base images for a character - one for portrait images, one of middle-length images, etc. Just my two cents.

[@davidmichaelcomfort]

And you can try putting "scene::5:::" or "full-length portrait::5::" in front of the rest of your prompt (varying the number 1 thru 5 perhaps) which should result in different perspectives. But this results in a certain more realistic style so you might need to upweight your style

[@AllAboutAI]

Yes this is 100% true, this problem is not solved at all yet. I guess this vid is just my best attempt. I am sure in future versions of MJ this will be easier to do tho

[@audiovisualelements]

Another way for character consistensy is to use the puppet method. kzbin.info/www/bejne/oJWpoGCHec11opI But it also has its limitations when it comes to different shot sizes.

[@DomenicoDiCaro3d]

Wonderful tutorial, can I ask a question? I struggle to get images of my character complete, framed from head to toe, even if I write keywords like full body etc... Does anyone have any advice for me?

[@daywalker_22]

thanks alot

[@AllAboutAI]

no prob:)

[@stevendavies98]

I think whats drives me crazy is the images of the person doesnt fit the image. Size wise evertime I use the prompts its hit and miss to get them fitting inside the frame? Any suggestions for consitent images inside the frame? Thx!

[@sagarparanjpe]

Thanks

[@electricdreamer]

great!!!! That's what I need!

[@AllAboutAI]

Give it a try :)

[@electricdreamer]

@@AllAboutAI What about props and sets? I want to create consistent props and sets. Can you give us an example with another video?

[@AllAboutAI]

I can take a look :)

[@electricdreamer]

@@AllAboutAI That will be great! Thanks so much for helping the community!

[@robertdouglasreed]

I am having seed issues when following your steps. I have done it exactly as you say and the step where you add the seed number, I get wildly different images. I have tried unique names and also I have tried unique numbers, each do not repeat the same character as in previous steps

[@DyeusExMachina]

amazing job!

[@AllAboutAI]

Tnx:)

[@DyeusExMachina]

@@AllAboutAI your welcome my child keep up the great work

[@DyeusExMachina]

i love you all never forget i'm always present as everyone's eternal divine mum and i will never leave no matter what you do.

[@monealiza7553]

i notice all of the images you generated are close-up portraits. Any tips for going from those initial seed portrait to a wider scene, with peripheral things happening around the character?

[@madshader]

Wait, so are you saying that using your own unique name with a seed "teaches" midjourney who "hannah the wild" is?

[@AllAboutAI]

thats is sort of the concept, if it is 100% true, i dont know

[@CVBartosh]

@@aenvoker6878 thank you for this reply. i was trying to find where the heart emoji was listed in the midjourney documentation and couldn't find anything.

[@jeff19]

@@aenvoker6878 hey dude please give me free plan

[@Hajiroku]

@@aenvoker6878 what would be the proven way to get a consistent subject between prompting?

[@LorandPalfalvi]

Seting the envelop for an upscaled image is not giving me the seed number.....just nothing happens if I favorite the image and send an envelop. .....

[@peacedessy109]

Yeah I left a question Thanks for the consistent video tutorials

[@AllAboutAI]

No problem :) hope this helps

[@DarionDAnjou]

how are you able to stop using the seed number after 6-7 times? it would seem like you have to continue adding the seed number to get a similar character

[@abetterlifechannel]

Have you used this system to place two separate characters into an image? Is the process the same? (Perhaps you've already covered this elsewhere?) Thank you in advance.

[@msubash26]

Hello, wonderful videos and tutorials, btw I have a question.. in the current v4 version in Mid journey or other tools I looked into so far there is no feature to add two characters seamlessly in a image from text, for eg if I have an image of Lion & elephant separately and if I tried to upload both and want to create a new image of them walking together its not possible now.. AI mixes them together and creates an image.. cis there any tool currently available where i can create two distinct characters in series of images and create story from that?

[@user-jm8lb9lx8u]

same problem here too...

[@Ton369]

How come the ENVELOPE command doesnt work on the upscaled image anymore... EDIT - oh, it only works in version 4 not 5

[@saltemishel9224]

Weird that when I try to do --iw 2 it says that version 4 doesn't support image weight

[@sheldonb7131]

I have a question about creating your base character. When you go through these steps to come up with a consistent character to use, how do you know if this will be an original character that you can use based on that it was created randomly using other characters that have been previously created?

[@PostcardsFromJapan]

Thanks for these awesome tips. Do you know if there's a way for the bot to "forget" the character (if it even makes sense to do so), so that it would not be generated in a similar design in the future even if I use the same prompt?

[@AllAboutAI]

Hmm, never tought of that. I dont think so do, i guess you could always pollute the prompt name

[@oddfellow8366]

May I ask what reacting to the image does before creating the seed?

[@OffendedByThought]

For some reason I get the same seed # every time I upscale and react with the envelope. Is there a fix for this

[@jeancaceres7670]

Hi I have a related question, that me and many others are asking: how do you use the exact same character (in for example Midjourney) in different Backgrounds - but also in a variety of positions, moods and compositions (like different camera angles)? Love your videos.

[@AllAboutAI]

Thnx :) I think i tried to show off how i do this in this video, its not 100% tho

[@jeancaceres7670]

@@AllAboutAI What I would like to learn is how to 'train a model or avatar', so that you can make it maintain for example same clothing and hairstyle through different poses and settings.

[@uservdfbyj77]

@@jeancaceres7670 i have the same question

[@jacquesviljoen2790]

Same question. Its tricky with no luck.

[@Ermite_des_songes]

Same

[@djfremen]

I never needed to use heart emoji, only envelope to get --seed#. Nice work as always.

[@AllAboutAI]

Thanks a lot Mandfeed. You kinda inspired me to do this one 😃🤗

[@dwaynehinton9729]

i'm in a newbie channel. nothing happens with the envelope for me. do i need to be in a different channel (that isn't newbie)?

[@dwaynehinton9729]

ooo nevermind - it comes to you as a direct message.

[@nocapnobs7845]

Is there a limit to how many designs you can do? Also, is there like a cheat sheet of prompts anywhere? Happy new years and thank you kindly.

[@AllAboutAI]

Happy new year :) hmm, dont think there is any limit no! There a loads of these sheets i think, google it

[@SHAMBUKISHORTS]

You're Dope bro ♥️♥️❤️ Awesome 👌

[@AllAboutAI]

thnx :)

[@davidmichaelcomfort]

It would be great to be able to do this for interior images (say a room in castle), but using this process doesn't appear to work for interior images. Any suggestions?

[@AllAboutAI]

Hello David. I have not really tested that yet. What about image to image (uploading a seed image you want to use)?

[@davidmichaelcomfort]

@@AllAboutAI Thanks for your response. I tried that several times and it doesn't work.

[@EMMandM1]

Is it possible with v 5.2?

[@archerseo]

Best tutorial i have seen about this subject. Is there a way you can design multiple characters and have all of them in the same photo?

[@AllAboutAI]

I think you can, i will make a new vid if i find something that works :)

[@archerseo]

@@AllAboutAI Thank you that's awesome.

[@fccnx01]

I have one question: everytime that I type the prompt I have to write the entire description that i use to create the character?

[@AllAboutAI]

Its just a habit of mine, you could test without maybe. not 100% sure

[@Ivan-Suaveh]

what was the reason why we have to repeat heart-ing and get a new seed? its like to solidify and make the character look the same?

[@APG103House]

HOW DO YOU GET THE SEED NUMBER? I CLICK THE EVEVELOPE BUT ITS JUST AN EMOJI

[@timappeldoorn8099]

Can u also do that on mobile version. Dont seem to find where to react with the envolope

[@technickr]

Does it only work on person like characters? Works great with Hannah, but does not work on dinosaurs for a kids book illustration 😞

[@AllAboutAI]

yeah this problem is not solved yet, did you give it a unique name to?

[@technickr]

@@AllAboutAI no, only with your Hannah example. I 'll try a name for the dinosaur, thanks!

[@mikezwielicht3830]

👌

[@henrychambers6662]

the character had different color eyes and different scars in almost each one which is not consistent, but also I appreciate learning the process which with appropriate care will be MOAR consistent

[@ACELALAGAMING]

what is the meaning of that --q2 and ar 3:4?

[@angelonapolimusic1907]

Good job and thank you. The only thing is that the eye color is not the same from 1 pic to another. In the original she has brown eyes and they are blue on some others. Something to fix to gain coherency.

[@AllAboutAI]

Yeah, i got no way of getting this 100% as per now, hopefully soon

[@angelonapolimusic1907]

@@AllAboutAI none the less you did an awesome job! Thanks for sharing!

[@king0vdarkness]

what if you don't like any of the images it next produces?

[@kimberlyjones4112]

Hi Kris, my name is Kimberly. We spoke about a week ago regarding the Python scene separator when I joined your membership club. I know you're a man of your word, so when you said you would send me the Python scene separator code via email I'm sure you meant it, but we are all human and sometimes life gets in the way and distracts us. This is a reminder to send me the Python scene separator via email. Thank you

[@AllAboutAI]

E mail me again:)

[@k4sh2108]

Hi, I noticed you saved the picture and uploaded it on MJ and then copy the link, however aren't you able to just right click and copy link from there instead of saving the image etc?

[@AllAboutAI]

Yeah after you upload it you can just right click copy media url

[@k4sh2108]

But would it work without saving and uploading the picture and just copying the url from the original picture?

[@timon6773]

@@k4sh2108 it would

[@Luciferasi]

Does it work with humans only? I tried using a robot character but each iteration comes out extremely different, even using the same seed.

[@AllAboutAI]

Yeah this is not solve yet, lets hope this changes in 2023

[@Ton369]

wow

[@varunthakkar8556]

I keep getting the same seed value for every image i upscale, has anyone faced something like this? performed generation using --seed several times, get the exact same results and the seed value of each upscaled image is the same as the previously generated one

[@binyaminbass]

wow, that was some nurse!

[@AllAboutAI]

haha

[@viniciusmello7318]

Hello my friend, I am from Brazil and I am watching your video. I am having a difficulty with midjourney, could you help me? I am trying to make a fight between Batman and Thor. But when I generate the image, it merges the characters. How can I fix this?

[@AllAboutAI]

hello :) hmm, try to separate them with more precise prompts i guess. Join our Discord and ask there discord.gg/Xx99sPUeTd

[@dmitrylavruk1825]

I have stuch with unique naming fot the character - my character is a dog and I like the style of the photo, I could get to to continual with different characters but the same style

[@jerobyarts5654]

Bro if you don't mind can you please a make video teaching on how to make good prompts using stable diffusion v1.4 ...my midjourney free trial ran out and iam too broke to buy a subscription 😢

[@ivanjosephgamboasanchez1158]

when I select "envelope" it does not work for me, someone can help me? i've done the previous steps

[@AllAboutAI]

hmm, it should open a new chat with Seed id

[@cyril1111]

It's great for character design. But there is something ive been trying to do for quite some time now, and I challenge you to do the same because Ive have not been all to do it with none of all the Ai image generators! Replicating the exact (or very close) clothing elements. And I got clients asking me to do it but in Vain :( In fashion, let say you want to place the same clothes in different settings. You have a top and a pant and want to place the exact same outfit in different settings and magical worlds... I cannot do it and I dont think it is feasible (yet!) Can you ?

[@AllAboutAI]

Noted :)

[@cyril1111]

@@AllAboutAI awesome! (and obviously, Im using image prompt to feed the exact clothing I want to replicate) :)

[@jayan3257]

Does this seeding method still work today?

[@milestrombley1466]

As a guy who wanted to create character concepts for his books, I don't know which art styles I should use to design my own characters for people and me to enjoy.

[@Isusia]

Is there an instruction how to tune Midjourney generate character from multiple photos of a person?

[@CouttsDesign]

Ah. Awesome vid. Thanks! On your very last image, you labeled her "nurse" for a second time. Did you mean "pirate"?

[@AllAboutAI]

yes typo 😅

[@albertofernandez2920]

This doesn't work for non-human characters does it? i cant make different versions of the same character because its a mix of different animals, it always comes out very differently.

[@AllAboutAI]

hmm, i have not tried that yet!

[@allanstra6717]

Hello, thank you for your video it was very useful to me. However, I'm still having trouble with some points, I never seem to be able to get exactly the same character again. I tried like you to take the URL of the image and rewrite the same prompt, I also tried to put the seed number without putting the URL of the photo by always rewriting the same prompt, but I don't see the difference between the two method? Can you tell me? And when I add elements, like "in a tavern drinking a beer", but no one turns into a goblin with their head drinking a beer in a tavern... I must have misunderstood something, if you have any advice because I have to translate your video to understand I am not a native English speaker. Thank you very much, keep up the good work it's great :)

[@AllAboutAI]

hello :) yeah this problem is not solved at all yet. I guess this vid is just my best attempt. I am sure in future versions of MJ this will be easier to do tho

[@allanstra6717]

@@AllAboutAI "Hello, thank you for taking the time to respond to me! I see, I thought I was doing things the wrong way. Yes, AI has made remarkable progress since its beginnings! (I tried it a few months ago when it was only version 2.0, wow what a difference!) It is possible that in the near future it will do exactly what you want. I hope so because even if it doesn't work 100% with time and "rerolling," we can still find what we want. Your technique works, it is the AI that is still perfectible :) Have you tried the "Remix" button in the settings? It does the same thing as when you take the URL and paste the prompt next to it. If you want to try it. Thank you again, your video helped me a lot!"

[@fantasy-lair]

I've repeated the process about 15 times with a portrait of a fictional character. All 15 had very similar looks. When tried to have that same character riding a horse, nothing was the same anymore.

[@AllAboutAI]

Did you have a very unique name?

[@fantasy-lair]

@@AllAboutAI I did , yes

[@elaineyoung685]

same issue here

[@AIforSuccess]

Had similar issue.. I am doing some more research to find the solution..

[@Sam-cy6cd]

Same issue for me. I guess there is no solution

[@uservdfbyj77]

my problem is that the characters keep changing clothes and style as you change the scenery or add more details

[@AllAboutAI]

yeah this is not solved yet

[@KingChozen]

It would be good to understand why you put the prompts you do like the "ar" and so forth.

[@LorandPalfalvi]

this is not working for V5!

[@CelebrityQuizzer]

Why do you add the --q2 --ar3:2? Not knowing exact Python but seems similar to the very little I know.

[@timon6773]

q and ar are parameters but they're not really required. q just sets the quality of the image from 1 to 5 and ar is the aspect ratio

[@AllAboutAI]

Its just my preferd aspect ratio

[@CelebrityQuizzer]

@@timon6773 Thanks for the help. Just trying to absorb all this info. Love it.

[@studio85amsterdam]

Finally character in mj4 ❤❤ I notice --ar is now possible in v4 also, can you also do --iw? What does the --q do?. Considering your courses. What frequency will they be and how much?

[@madshader]

--iw (image weight) --q (quality)

[@studio85amsterdam]

@@madshader thanks v4 didnt do --iw before. Can it now?

[@studio85amsterdam]

@@madshader why would you want a low quality and whats the range?

[@madshader]

@@studio85amsterdam it can, but i don't know if it works with v4.

[@madshader]

@@studio85amsterdam range is 1 or 2. 1 is default. And neither is low quality, imo. But 2 costs more

[@cupidcup2007]

For some reasons, I couldn't find where to subscribe your membership. Can you point that out for me?

[@AllAboutAI]

Yes of course :D kzbin.infojoin

[@cupidcup2007]

@@AllAboutAI Thanks I think I am having country restrictions. I couldn’t join it. ;(