HUGE Update: Create Consistent Characters in Midjourney (Easy Tutorial)
ฝัง
- เผยแพร่เมื่อ 8 พ.ค. 2024
- In this video tutorial, I'll show you how to create consistent characters inside Midjourney.
The Code used in the Tutorial:
--cref (image link)
--cw 0-100
0 = More Creativity
100 = More Consistency
Join Our Newsletter for the Latest AI Filmmaking News: curiousrefuge.com/
Our AI Filmmaking Course Starts Soon: curiousrefuge.com/ai-filmmaking
Check Out AI Advertising: curiousrefuge.com/ai-advertising
0:00 Welcome
0:12 Getting Started
1:43 Breaking Down the Code
2:35 My Favorite Midjourney Hack
3:27 More Creativity with Consistency
3:57 Tips for Different Clothes
4:52 Creating Myself
5:46 An Important Lighting Tip
6:38 My Final Thoughts - บันเทิง
Finallyyyy !!! it used to take ages to train your own consistent character. This is just what we've been waiting for ! Thanks
No problem! Can't wait to see what you create next!
Excellent tutorial! Thanks for another great video!
Thank you for the short video, straight to the point, and prompt in the description.
Our pleasure!
Great Video, thank you Curious. Results are great.
OMG finally! I have been waiting for this one
We're so excited too!
This is what I need !! More like this please!!!
Finally! Thanks for the heads up, I am going straight to MJ to try it.
Finally! Thanks for the walk-through.
You bet! More to come!
Good explanation ! Thank you !
Clear and straight to the point. It didn't work as expected this time but I'm sure It will in the future!
Really appreciate, you're amazing, it works and the results are fantastic! 🙏 (Subscribed)
Awesome, thank you!
thank you for posting this... very helpful
You're very welcome!
Thank you dude! Was very helpful
Glad it helped!
This is such good news!!!
Very nice feature 😊✨
Nice!! Both Midjourney and Caleb have done a great job.
We appreciate it!
FAN..... TASTIC!!!!!!!!!!! Jewelry consistency and prop consistency are still likely to be tricky, but it's a HUGE step forwards for AI filmmakers!!
True, it's those little details like jewelry that can really help sell a scene. I'm sure those will be coming up soon but for now we will likely rely on inpainting and photoshop :)
Yeah, we need to see how this does. Generating training data for stable diffusion
i dont see the "open in browser" option underneath the image i just uploaded though?
Lmao
Superb instructional video. Thank you! Worth checking out also prettyface ai for consistent characters. Less options compared to MidJourney though
Such big news, such great news! 🎉🎉🎉🎉
Right?! Lots of interesting things happening this year.
@@curiousrefuge for sure and while I'm here, a big thanks for this video it was really useful.
What?!!! Even more character consistency beyond -sref?!! This is great!! Cheers 🍻
Right?! More and more improvements each month :)
Great - this is what I need:)
YOU FOOKIN EARN MY SUBSCRIPTION! GOOD JOB MATE! KEEP GOINGGGGGG
I am signing up for the course. Thanks for the uploads
Heck yeah! Excited to have you in the course. :)
Huge implications for sure -- love it.
Yes we're so excited!
Very helpful, liked and subscribed 🌴
Awesome, thank you!
Amazing, thank you for this easy to follow tutorial! I tried it and it worked! I am hoping to use midjourney AI models to produce images for my scarf brand. Consistent characters will be a huge help.
So glad it helped!
AMAZING!!!!!!!! thank you🙏
You're so welcome, Chen!
Really useful. Thank you.
You're welcome!
YES!!!! 🎉🎉🎉
We're excited too! One more push into the future :)
This is exciting.
We're excited too!
Nice! Thanks!
No problem!
Love it Caleb! Just about finished with my first weeks assignment and I was wondering. I saw in another video where someone uses the prefer option to lock in the character style. Have you done any playing with that?
That's a great idea! Yeah you can use that to lock in repeated characteristics for sure.
a video on image editing, product photography ❤
Cool, thank you. Glad to see Midjourney has made this update. Do you have any recommendations for creating the same character in a different scene, setting and or performing different actions...thank you.
We suggest testing using both sref, cref, and mixing with different prompts. You can use CREF on same characters but use your prompt to totally change the scenes.
GREAT VIDEO!!!!
Glad you enjoyed it
yes!!!!! that´s very huge update. jeeeeeeeeeeeeeee :D
Indeed it is! We're excited!
That was great. Thanks for sharing What is your suggestion for giving this picture some move and lips syncing? Pika ?
Yeah Pika is probably your easiest bet at the moment, but Wav2Lip 1.5.1 beta is great too from Synclabs too.
DAMN BRO thanks !
You're welcome!
Thanks!!
Our pleasure!
woooow amazing
Thanks for watching!
FINALLYYYY so tired of using option set and in painting
Wow!
In the words of Owen Wilson...."wow"...
I'm going to try to create a sprite sheet using this :D
Can't wait to see it!
Do itttt!!! Dont' forget to show us too :)
Huge news! Thanks! I wish you would have created pictures of yourself doing various things in various environments to show off consistency across a story board, which is what people need in order to create a consistent film. Like riding a horse, or running down a beach, sitting at a breakfast table, all with different actions, clothes, etc.
Also, can you have consistency with 2 or more characters in the same scene?
Good idea! As for the 2 character's, we'll do some more testing!
We’ve been waiting so long lol
Finally here :)
Alleluia
Exactly!
love how there were tools for consistency for stable diff before this 6 months ago. mid is slow but well
True! However, Midjourney is quite more accessible to the general population compared to SD.
@@curiousrefuge just cause it doesnt have that huge of a learning curve, if you ask me SD is accessible locally and if you dont have a gpu its also free on many sites
thanks for your tutorial, super good and easy to understand.
Glad it helped!
Hi! thanks for the tutorial, you deserve like and subscribe!
one question can we put together two consistent caharacter into 1 picture? let's say the pic of the girl and your photo in the same picture.
thanks !!
Yes! You can make your own parameter for each character, and then inpaint different character faces :)
You can just right click, copy link the image (no need to open in browser).
Great tip!
Thanks for the video! :)@@curiousrefuge
Thank you! I followed your prompt and used asian photos, and I realized the face does not look like the original image at all. I guess the new updates are most effective for white faces ;)
We'll have to do more testing! Feel free to show us your work in our Discord channel and we'll be happy to help!
How can we do this with landscapes? I would like to do a children's storybook, I have developed a scene in mid journey that I really like, countryside thay has a little stone Cottage, little picket fence, etc, I would like to do different angles from the same garden. Any tips?
Love this video by the way. I can work on my characters now for my book too!!
For landscape shots, definitely make sure that is added to your normal prompt. You can always inpaint over the face of a charcter using the --cref too if needed!
Can I use this for product photography? It would be great if you could do a video
Consistent objects will require some more editing (likely in photoshop) but you certainly could use it for that use case.
Thanks for the instruction, which plan do you have on Midjourney, because I think with the basic paid plan, you can not do this, am I right?
Yes, i dont think its possible at all, been trying for hours, the original person does not stay the same.
You can use parameters with any of these plans. Are you unable to use the --cref paramter?
Great vid. Much wanted feature. Anyone else entering the Runway film comp? I’m entering a book trailer. This is moving so incredibly fast - some of the methods I used are already outdated.
Good luck on the competition!
@@curiousrefuge Thanks!
Cool! Can we also have shots with multiple people on it?
That may require a bit more testing and inpainting :)
The thing that we need is pretty simple 1) Any image of an individual changed to reflect the MJ model. So that way you can pose and have whatever camera angle reflected in the final image. MJ is largely useless without it. 2) We need to be able to set a stage using collage or 3D assets or actual pictures and have MJ reflect that in the final image.
Ai has a long, long way to go.
True, but the advancement are moving fast so we are excited for the near future :)
Hallelujah! Can you use this for settings? Say I want characters to always be in the same kitchen in different scenes? What about multiple characters?
I havent tried myself but someone said you could use inpainting with it so you could inpaint a second character I think
We suggest adjusting your prompt to have a different setting. It shoudl still work (i.e., whether you say office room or beach). For different character in same scene, we will have to do some more testing!
really impressive, wonder what the underlying approach is, some sort of IPAdapter?
Good question! Not sure but we'll see if we can find out from their Discord server.
Is there a free to use talking avatar AI currently on the market or are the best models still D-ID or HeyGen?
You can try Wav2Lip! or Lalamu!
@@curiousrefuge thanks so much !
please tell us the best way to compine this with consistent style to make a story. thank you.
I will try! For consistent scenes you should definitely look at the "--sref" and for character it's the "--cref". As usual, certainly make sure you have prompts that have matching qualities too.
I have not decided on Midjourney yet (lots of choices). I have an image in another program that does not allow you to see the seeds or any address so, if I have Midjourney, I did not use a name and the prompt is specific but not with the details I want now (I made it in my early days on AI). I have one image that has two characters that I want to use more than any other image I have created. What do you advise if I want to change the backgrounds and can Midjourney change the stance of characters or do you have to start from scratch and try to get something similar but with names and character poses etc..? I am a nurse so, my tech skills are rusty and I am really at the beginning (maybe dabbled for a few months randomly here and there for an hour or two a week (if that).
I'd suggest jumping in on our Discord server and showing us the the exact prompts and generations. Our community would be more than happy to help!
Now with this update. Could you make a new tutorial on how to make a short trailer or movie?
That's exactly what we teach in our course! We'd love to have you join us. More info at curiousrefuge.com . Enrollment is open for just one more day!
Hi! is it possible to use this command with 2 characters?
Thanks!
Well need to do a bit more testing to confirm the results!
@@curiousrefuge I tried but got the same character twice :( still testing tho
Is it possible to put 2 characters in a single composition using this technique?
Yes - but it takes a bit more work. Try using inpainting with --cref ! Check out our discord and we can show you walk through!
before running your own midjourney server did you pay for the program? I imagine so but just wanted to ask incase you know something I dont. Also, I have created a server in discord but not one for midjourney. Is the process the same? Or how do I link my midjourney to my own server? Could you show us how to do that or point me to a vid where this process is explained please?
Good idea for a quick tutorial. If you already subscribe to a MJ subscription, then you can simply invite the MJ discord bot to a new server that you create in Discord. You can do this multiple times as the MJ bot is tied to the same subscription.
AH thank you for the reply. Im pretty sure i can figure out how to do it but i think it would be a good tutorial for sure for many@@curiousrefuge ..All the best!
Man, Appreciate that hustle! YOU have won the chicken dinner...for...like, a week!
::cue Rick Ross::
th-cam.com/video/gEVLvBRB96Y/w-d-xo.html
:)
Haha I was literally about to work on a course video about consistent characters and Dave texts me and says ‘have you seen this new consistent character tool’. I think I audibly gasped. 😂
@@curiousrefuge Haha, it's always the way! I had a plan to do a thing...and then, "Welp! There goes that plan!"
I swear, I'll come up with the best AI Video Workflow, and that afternoon Sora will drop!
I do the same way, but face not the same, can see the different of two different person, how can i improve it or you have other better way? tks.
Try getting multiple angles of the same character, then using it altogether in making a new parameter, and setting that parameter to the name of a character - then you can inpaint the characters faces.
@@curiousrefuge Very much thank you for your advice. Can you show me the sample or prompt too. appreciate.
What's the best way to do more than one character consistently using cref?
It would likely take some finessing within photoshop and also additional inpainting.
Now with consistent characters and consistent style in Midjourney Is there a way to make 30 consistent frames to make a shot that makes a video when combined? thank you.
We'd still suggest you use programs like Runway or Pika.
@@curiousrefuge thank you Caleb.
Would this work for a product? I have a hard time using my products in Midjourney. It never looks right
Unfortunately it works better characters and less for objects.
good
Thanks!
Is there a way of keeping a constant character but at different ages?
there isn't a parameter for aging but you can try inpainting wrinkles!
This a great video. Unfortunately, midjourny does not do mixed people. It gets confused.
Interesting! We'll have to do some more testing and see if there are ways to make it work.
Pair this with Sora's still-to-video functionality and it starts to look a lot like the AI cinema revolution we have been promised.
We're getting so close, aren't we!?
At least with my MJ bot you can just right click on the image after dragging into Discord, copy link, and use that -- you can skip the open in browser and copy step 👍
TRUE! We should have mentioned that :)
Can you actually make consistency in costumes? In our projects especially, the character's costume is often more recognizible than their face
For costumes you can certainly adjust using inpainting, but we'd suggest also taking into photoshop for inpainting the finer details.
Can you do animated images this way?
Pixverse just released a consistent character tool so you may want to check that out :)
so my question is, how do I now keep the character consistent but change pose/background IE woman sitting in a kitchen chair vs walking across a city street?
You can try changing the prompt, and using inpainting with the --cref paramter to change the face.
There has been a change with MJ, the face gets distorded now even when you use --cw 100. Has to do with the update dealing with deepfakes
Interesting, we'll do more testing :)
It’s cool so far, my issue I’ve noticed is that the characters aren’t very expressive:/
True! More fine tuning will be needed for that.
I'd prefer to have an offline tool rather than using a server, but I appreciate the video nonetheless
An offline tool would be great (if it was super easy to use :) )
There's others, hopefully we get one similar to the video soon@@curiousrefuge
Okay but how do you integrate the character into an compelling consistent environment? And how do you also change her facial expressions?
There are lots of other tricks you can use. For the facial expression and environment, we'd suggest also adjust the prompt. You can also do some inpainting!
great video but when i upload my image into discord there is no option to "open in browser" for me
thus im stuck at first hurdle
ps tried in a few different browsers
Make sure when you upload to discord, you press Enter/Return on keyboard to make sure it's actually in discord. Otherwise, when you right click you will simply see your computer's options rather than the discord options
@@curiousrefuge ahhh yes. I needed to press enter 3 times
Weird, when I click on the pic I can't see the option 'open in browser'? I can't do even the first step from the tutorial lol
Make sure when you paste into Discord you press Enter/Return on keyboard to make sure it's in there. Once it's in there, you will see all the options when you right click on it.
What if the photo you want to use doesnt have a link connected to it
You can paste into discord and once it is in discord you can follow the same process by right clicking and copy link :)
After uploading a MJ created image, I'm not getting the "open in browser link". All other images have the link, just not the ones I upload. Any idea how to fix this?
If you click on the image (so it expands a bit), then do you see the "open in browser" on the bottom left of the image?
No, there is no link after I upload images. I found a workaround though. The image was originally created in MJ. I right clicked and selected copy link and it worked. Thanks for the fast reply!
@@curiousrefuge
Does this work for animated characters?
We'll have to conduct some more testing!
When I am giving the command It shows "Please check that your URL is a direct link to an image, not a webpage.",kindly help
Instead of taking the direct URL trying right clicking and pressing "Copy Link" from discord on the image.
I did this too, u try again, i think its error from midjourney only, coz yesterday it was working
It doesn't work as well as I had hoped. Seems that it can't create consistent variations of illustrated characters and there are some weird glitches. Perhaps I need to test it a bit more and use more detailed prompts.
Interesting! We will do some more testing on our side too :)
mine is not working, i did exactly what you said and typed in but its giving me different characters and now not even listening to the prompt
Feel free to jump in our discord and let us know what the issue is!
Ffffffffiiiiiiiiiinalllyyyyyy!
Exactly! We're so pumped for this feature.
When I try to use --r 10, It prompts me asks if I am sure I want to want to imagine 10 prompts. I choose yes and then get an error that says "Too many prompts, this would expand into too many prompts, the limit is 4).
I am a paid subscriber. Why am I getting this error?
It’s because of your tier. The bigger the tier the more repeated prompts you can use.
Doesnt work. My prompt was : looking to his left, --cref "URL" --cw 100 --style raw --s 250 The result was nothing like the image in the url!! I just wanted the same character but looking left couldnt even do that!
Hmmm, let us know in our discord what image and prompts and we'll try and help!
@@curiousrefuge Thanks, I cant find the link to your Discord could you please let me know
What happens to seed no, we don't need that anymore ?
No, you could probably still use seed number for composition.
What a bit having two consistent characters in a prompt?
You can inpaint using --cref! It works well :)
I get an error message all the time saying the --cw 100 parameter is invalid
Typically this happens when the "copy link" you copied wasn't from discord but rather your computer. Make sure you are grabbing the right link :)
As 3D artist… it’s look so rough right now lol . I’m switching to small business. Give me grave god
Nooo it will enable you to create even more, faster! You got this!
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
Yes! Come to our discord and we can show you how to achieve this. But for now, you can use the --cref and inpaint the same characters once you have saved new characters as their own parameters.
copy image is an error, cr ar give error. also.
Which error? Are you typing "--cref" ?