How to Create Multiple Consistent Characters in Midjourney V6!
Vložit
- čas přidán 22. 05. 2024
- Here's a simple and quick tutorial on how to create multiple consistent characters in a scene with Midjourney's new --cref feature. This is the most common question I get, so I did some testing to see what Midjourney can do.
We'll try generating multiple consistent characters with different camera angles, character interactions, backgrounds & environments.
FREE PDF Prompt Guides, Tutorials, etc:
taoprompts.gumroad.com/
My Instagram:
/ taoprompts
Chapters:
00:00 Multiple Consistent Characters in Midjourney
00:31 Create Reference Photos
02:00 Generating Base Background Photo
02:44 Adding Consistent Characters into Photo
05:54 Getting Consistent Hairstyle
08:08 Testing Anime Style Images
#midjourneyprompts #midjourney #midjourneyv6 #midjourneytutorial #aiartprompts #aiarttutorial #midjourneytutorial #aiphotography
I got so many requests for this video!
If you want to learn more about prompting for separate characters inside the same image (clothing, background, colors, etc), here's a video guide: czcams.com/video/F53ti2sD7KI/video.html
Tao. Unfortunately, my subscription for midjourney is over and I can't afford 30 dollars a month. could you please make the same awesome tutorials about consistent characters, consistent style and good promoting but in Dalle 3 inside Microsoft Bing chat which is free because I couldn't find any CZcamsr teach about this and there are millions like me who can't afford expensive tools. thank you.
Hey, I think Midjourney has a cheaper plan at $10 a month. Do you find that you need more GPU time than that plan provides?@@AItechIsBeautiful
@@taoprompts yeah , I took the 10 dollars plan it lasted with me only 10 days.
I've seen people do consistent characters with DallE + Chatgpt, but I haven't seen anything that lets you do it with just DallE and the bing image creator unfortunately.@@AItechIsBeautiful
@@taoprompts yeah. thank you Tao.
Amazing tutorial Tao!😀
Very well explained. Love it!
Fantastic video my man.
Thank you for your chill and knowledgeable insight.
Sending chill vibes your way 👍
Tao, thank You so much for all videos ! Your explanation is always very clear and simple :) really appreciate that
Of course! Midjourney is complicated so I try to make it as simple as possible.
as usual, epic share & very clear descriptions of the steps
Thank you, multiple consistent characters still needs some work, but the potential is definitely there.
@@taoprompts for me, I didn't realize you can inject cref when doing inpainting -- that's really useful '-)
Dude, thank you for everything.
Great Video - all clear and well done explained! thank you so much for making it and making our life’s in MDj better ❤
Thanks! Midjourney 6.1 should be just around the corner so a lot of exciting updates coming soon 🙏
it takes a lot of time and effort to explain it so structured and simple! well done and thank you :)
Thanks Bibi, these do take sometime, but I have a lot of fun making them. I'm glad you find them helpful👍
Awesome method!!
Thank you! This was incredibly helpful!
You're welcome 🙏! Glad to hear this was helpful.
Amazing tutorial as always ❤
Thanks Stevie 👍
Very good video, carefully explained and demonstrated. Subscribed 👍🏾
Thanks for subbing! I've got a lot more guides coming soon.
@@taoprompts more niji stuff would be great 👍🏾
Great Tutorial Tao, will use this method.
Thanks Dani!
thanks bro man...ur a gem.......just what im looking for..!!!😀.but it looks like we are little more close to our goal.............very good work.
Thanks man 👍. Midjourney is getting really good, I'm honestly very surprised it was able to get all the different camera angles and face directions correct.
Great video! I've been saying this for several months now, AI-powered storytelling is the new gold rush for creatives! I think everyone who has a story in their heads should at least give it a go. You owe yourself that much!
Absolutely, with image, video, and audio Ai, a lot is possible nowadays 👍
great video! im learning a lot with your content. im using this for my meme page. greetings from Peru
That's awesome! I've been seeing a lot of Ai memes nowadays. Hello to Peru🖐️.
Well done. Great work.
Thanks man, these videos are lots of fun for me.
As usual your video is excellent!
Thanks man! This was an exciting update.
@@taoprompts Yes! What the best for your channel for the future
Wonderful! You accepted the challenge. You are the Midjourney Maestro, Tao. Hopefully, they'll make it easier in the near future.
Thank you! I just spend a lot of time testing stuff out.
They've been really responsive to community feedback, so if we make a lot of noise about multiple characters, I'm sure they will consider adding in a feature to make it easier.
Easier? It's already easy af. Perhaps I could help if you're having issues with the --cref process
Dude you're a lifesaver!
I got you man!
Yessss The MJ prompt master schools us Again!! 🎉✨👏🏻💯
Thanks Stylist 👊, it's always great to see you enjoying these videos!
thank you so much for your videos
Absolutely man, I'm glad you like them 🫵
Awesome Great video and informative! thank you!
Thanks Val, always great to see you in the comments!
@@taoprompts you’re awesome!
Yea you're really good. I subscribed
Thanks for the sub! I got many more guides planned soon 👍
Great tutorial. I am scratching my head these days 😁
Thanks man! I will try to make it easy to understand 👍
glad to be the 1st comment! I like to thank you for your material really has been helping me a lot. what do you recommend for building a comic book style using MJ. and its new consistent characters. thanks
Thanks man! Do you mean how to create comic strips?
There's a channel called tokenized ai that has a great video on creating comics called "I Created a Midjourney COMIC BOOK: "Treasures Beyond Gold"
You could combine what he did with the --cref feature to get comics with consistent characters.
1st off thank you for the "Prompt" response,,U see what I did there ;) "Prompt"ok so yeah I will most deft. be checking that out...thanks and keep them coming! I know your channel will blow up ....trust me it will :)@@taoprompts
Haha, thanks man. I will test out some cartoon and anime prompts myself also.@@safetyhomeloans950
Thanks Tao, you rule
Appreciate the walkthrough. I originally came up with a process that cuts this in half. Just prompt for Lisa (or whoever has the more difficult hairstyle) and an Asian Man in blue shirt etc, then just only inpaint kim. This worked as well but if they are opposite skin tones as in your examples, it may become an issue with extra inpainting being needed.
That would certainly save some time, but yeah if you use a single character reference for 2 people, they usually both get the same skin tone and facial features as the character reference, so you'll need to do some more inpainting.
But on the other hand you can get higher quality for the first character reference, since inpainting often gives slightly lower quality as far as I can tell.
@@taoprompts yes I agree. I usually prompt for the woman as that's the one most people are looking lol and then in paint the man
That's a good strategy, I do wish they would gives us an option to increase the inpainting quality though.@@BabylonBaller
You are the man!
Thanks man, hope this guide was helpful 👍
very useful
Your tutorials are great as always. I hope Midjourney will make it easier to pose multiple consistent characters together, without inpainting being needed. Hopefully, Midjourney will not only create consistent faces, but also consistent hair & clothing. Also, I wish I could upload an image of a pose in Midjourney, and ask Midjourney to put my consistent character in that pose. When it comes to character posing, I’d also like Midjourney to copy more complex poses of characters interacting with each other (e.g., a pose of two of my consistent characters hugging each other).
Thanks 👍. To get really fine-grained control over the pose you need to use something like ControlNet. Midjourney doesn't have that available but Leonardo Ai does and it's worth trying out.
Also fooocus, u can download on it on your computer. Its not that difficult and once set up its quite good
very cool
Fabulous!!! Why do you not have 100,000 likes and comments!!
Thank you! I hope that happens someday 👍
Brilliant ❤❤❤
Glad you liked it! 🫵
great video
Thank you!
I liked the advanced nature and tips! The Anime looked like a good match to me but missing those green triangles on dudes face, IDK. Anyways was looking to see what you thought on Multi-Characters, predicted right but enjoyed a good amount of tips in your workflow! Nice work!
Thanks man! Anime is so stylized, there's so many different ways you can draw an anime face. I can see how the inpainted man is same as the original character.
It’s not just the green triangles that are missing. The body types of the generated images just don’t match the original character references. Based on the reference image, the male character’s shoulders should be a lot broader than they are on the generated images. The female character also seems to have a different body type if you look more closely.
great video thank you for creating this, my only question is when you are creating the character are you assigning -cw within that master prompt before you create the prefix? Or no?
I'm not adding --cw in there since by default whenever you use --cref the character weight. --cw is set to 100. But you can certainly try adding that in and changing the weight if you want.
By default whenever you use --cref the character weight --cw is automatically set to 100. I didn't add that into the prompt, but you can try changing up the character weight to something less than 100 though.
I love brother!
Very useful! What prompts can be used to adjust the direction of gaze and body orientation of the two people?
I would add in camera angles to the start of the prompts like:
"side angle shot from the side" "side profile shots"
Or prompt for specific activities where people would have different orientations like:
"playing a board game"
It is hard to control the direction they are looking and body orientation though.
Needed this! So you have to use --cref every time you use the saved character?
Yes, you'll need to use --cref whenever you want to use the consistent character feature.
A lot of people have been having issues with the --cref and vary region box recently. A workaround for now is to just directly use the reference image url instead of /prefer_option_set.
"--cref image_url"
@@taoprompts Ahhh, okay. I've been playing around with doing it that way (adding reference image + cref) and it's working great. Thank you sooo much. 🙏🏾
Thanks for sharing! Worth mentioning... No Ai company (MidJourney included) has figured out how to do this well for our OWN faces. Every time I do it, I get this weird looking dude (he always looks the same), and he definitely does not look like me. Faceswapper wasn't much better either.
To get the best results for generating your own faces you would need to finetune (train) a custom model. It can be done with Stable Diffusion + Lora/Easyphoto/InstantID, I've done it myself. Although that does require some extra setup on your end.
@@taoprompts Thanks man. Feels like a good future video for you...!
First of, props to your tutorials, appreciate it, brother. Could you please look in to this and provide me a solution, Error ('cref', expected at least one argument), please let me know how to fix this error. Thanks in advance.
Impressive, could adding the cw command increase even more the consistency of the characters ? Can you make a video with the command? Thanks!
Hey, when you're using consistent characters, the cw feature is by default set to the maximum value of 100. So, for this video all prompts by default also have --cw 100 added to them.
@@taoprompts oh I didn't know it was added by default ! Thanks
👌👌
Great tutorials! Thank you! When I try using 'vary region' with my saved character reference shortcut, I get the message ('cref', 'expected at least one argument') do you know what this might be? It works in a straight prompt.
Not sure what the problem is but I'm also having trouble getting it to work correctly.
For now, a workaround is to just directly use the character reference image url, instead of prefer option set inside the vary region box.
--cref image_url
@@taoprompts got it. And thanks again for your tutorials. Best I’ve found!
Hi and thank you for this great video! I have a problem when it comes to the point where you change the characters with "vary region" (03:20). I created the characters with „/prefer option set“ as you said before. I gave them a name with "option" and used a image-link with "value". But when it comes to the point where i want to change the characters with "vary (region)" I get an error message which says "Request cancelled due to image filters". Can you tell me about the problem I have here? I used exactly the prompt you used (just with different characters) and the parameters "--cref --Eva --ar 16:9 --style raw" Thank you in advance!
Hey, it sounds like Midjourney is generating images that are NSFW with your prompt & reference combinations. Sometimes this can occur even when trying to generate regular images. You could try using different character reference photos or switch up the prompt slightly add see if that fixes things.
@@taoprompts thanks for your message, i don't know what would be NSFW generating my prompt, due it's just showing regular people :D But i will try again with a new one. thank you!
Hello Tao!
Has anything changed in the "Vary (Region)" box?
trying to change the characters' faces, but my prompt command is refused.
Yes, I am also getting a similar error.
For now, a workaround is to just directly use the character reference image url, instead of prefer option set inside the vary region box.
--cref image_url
Great video now all they have to do is get the fingers right?
Yeah, fingers are still a struggle unfortunately. I haven't seen any Ai yet that can do hands well consistently yet.
Hi Tao Prompts, thanks a lot for the tutorial!
I keep getting this error message:
Error
('cref', 'expected at least one argument')
no matter what I do, the syntax appears to be wrong, I've tried everything from putting --cref --bob before and after the prompt (and everywhere else) and can't find any help online with this. Please help?
Me too! I have created lots of pictures using this method two days ago, but I can't use --cref --name today
@@yaningzhu7433 hey, thankfully I fixed it. For some reason, it only works with --cref (URL) it doesn't work when you type --cref -- (URL).
It's the '--' before the URL that needs to be removed. A moderator told me in discord. Maybe they updated it to work this way since this video? Who knows
I'm also having trouble getting it to work correctly.
For now, a workaround is to just directly use the character reference image url, instead of prefer option set inside the vary region box.
--cref image_url
I had no idea I could name characters and was just playing with --cref
/prefer_option_set is super useful, you can save custom parameters in addition to an image link.
@taoprompts ... I pay for second tier Midjourney ($24/month). I'm familiar with --cref and --sref, etc. but when i try this method you're showing on this video i get "--cref expected at least one argument" notification and nothing happens. Do you know why this is?
I'm in the "vary region", i mask her face (and hair, I've tried both), type --cref --(my ai models name) after the camera type and before the --s 750 --v 6.0... just like you did in the video. I get nothing. 💁🏽♂️🤷🏽♂️
Hey, a lot of people have been having issues with the --cref and vary region box recently. A workaround for now is to just directly use the reference image url instead of prefer option set.
"--cref image_url"
thanks, so . question, I need to create diferent ages, like 10 year old 20 year old 40 year old.. how can I do' I try view but do not work, maybe have to low the CW?
You could try using a base prompt like this one: "a photo of three people in a restaurant. on the left is a 10 year old boy, in the middle is a 20 year old college student, and on the right is a 40 year old man --ar 2:1 --style raw"
And then inpaint each of their faces with reference photos, each with a different age person.
I have same guy , so need to show him aging @@taoprompts
I keep getting 'Request cancelled due to image filters' as an error every time I try the lasso tool and try replacing the face :( Any fix to this?
It sounds like Midjourney thinks the images generated should be censored. It could be that the images are NSFW or have certain public figures in them, like the US presidential candidates which are banned atm.
can you get this methode for product? like perfume?
You can try using the cref feature for products, but it won't be able to replicate the original image exactly.
Can't we give some name to the product?
PLEASE ANSWER!!!
Why it doesn't work to me it always says error ('cref', 'expected at least one argument')
I get the error message as duplicate images are not allowed, what might be the reason
During which stage are you getting that error? That usually happens if you try to blend two of the same image together or something like that.
@@taoprompts thank you, error resolved
After I lasso and type the "--cref --[name]" prompt, I get this message "Error ('cref', 'expected at least one argument')". What does this mean, and how do I fix this?
I think a lot of people have had trouble with the vary region tool recently. For now a work around is to just directly paste in the image url link instead of using /prefer_option_set. s
So "--cref image_url_for_character"
Me to…that’s a good fix. Thanks
@@taoprompts Yeah I tried that, and it works! But it does not work as well as the one you showed us in the video though. Thanks anyway!
Hi Tao! I have a question, is that available for free Midjourney users?
I wasn't aware that Midjourney has a free trial, but you should be able to access all the features.
@@taoprompts thank you for your reply.
After posing multiple consistent characters together in Midjourney, can we then change the characters’ facial expressions & gaze?
I think you could probably try, although I'm not sure how well that would work. For controlling the gaze, you'd probably want to have that already done in the base image prompt before injecting the consistent characters into it.
How can u do pasted image Vary Region? So that u don't make extra generated images
I'm not sure I understand what you mean. If you use the vary region tool, Midjourney will always generate 4 different variations.
@@taoprompts I meant, you have a picture and you're going to post the picture in Discord midjourney. So, you want to vary a specific picture.
@@mzsutodraw There isn't any way to use vary region for your own uploaded images atm.
It appears that the reference image you use has to be from the same version of Midjourney. When I was using v5 after generating images I switched to v6 to try the --cref feature. The images produced had none of the key features of the referenced character. When I stayed in v6 and generated the images and then referenced them in different scenes you could see the consistency. Has anyone else experienced this?
Yes that would make sense, it's designed to work with images generated with the same model.
Is anyone else having the Error "'cref', 'expected at least one argument'" ?
or
does anyone knows what to do?
I'm getting the error "cref", 'expected at least one argument
Midjourney appears to have made some changes to the vary region box and that's messing up the /prefer_option_set command. A temporary workaround is to just attach the image_url directly after cref instead of using /prefer_option_set.
"--cref image_url_1 image_url_2"
is there any midjourney pro, who is experienced with multiple consistent characters in different styles? i need illustrations for different books. different detailed scenes with 1-3 consistent characters each. i pay for the projects. on upwork and fiverr they dont have any skill.. just answer on my comment if you think you can help and earn money with it
('cref', 'expected at least one argument') what is the meaning of this error?
Yep same here
Midjourney appears to have made some changes to the vary region box and that's messing up the /prefer_option_set command. A temporary workaround is to just attach the image_url directly after cref instead of using /prefer_option_set.
"--cref image_url_1 image_url_2 --cw 50"
Brother, could you please add subtitles in Turkish when adding automatic subtitles to the videos?
Sure thing. You can also use the auto translate feature in the subtitles/cc tab to translate the subtitles to any language you need.
--cref --name can not be used now, I don't know why. The Error said ('cref', 'expected at least one argument')
me too
Hey it does seem like people are having issues with --cref.
However, it does still seem to work if you just directly use the image url instead of prefer option set inside the vary region box:
--cref image_url
Midjourney is decent at adding a consistent character’s face to the image, but the drawback here is that the character’s hair & body type are usually off. The hair & body type doesn’t match the original characters. Also, the clothing looks close, but not quite the same. Clothing & hair style are mutable, so sometimes I want to keep them the same, and in other scenes I may want to change them up. However, the body type of a consistent character should remain the same in every scene.
It is hard to keep the face and body exactly the same. Midjourney will try to maintain the main features, but the details tend to be off.
How do i make my skinny character fat? no one seems to be giving me the answer.
You can try prompting for a fat character and then turn the character weight down to 0 "--cw 0". That seems to work okay for me.
Keep getting “cref expect at least one argument” on vary region
Midjourney appears to have made some changes to the vary region box and that's messing up the /prefer_option_set command. A temporary workaround is to just attach the image_url directly after cref instead of using /prefer_option_set.
"--cref image_url_1 image_url_2"
MJ Vary region not working on IOS at the moment and I use it for the face swap as a consistency character tool for my shorts. V frustrating. I hope they get it fixed soon.
is it possible to use midjourney for free?
There are only paid subscription plans available. The Bing image creator gives you some credits you can use everyday if you want to try that.
doesn't work with --name
There seems to be some issues with vary region and the cref feature right now.
For now, a workaround is to just directly use the character reference image url, instead of prefer option set inside the vary region box.
--cref image_url