Create Consistent Characters with Different Expressions in Midjourney!

[@taoprompts]

Here's an UPDATED video for the new --CREF Character Consistency feature in Midjourney: kzbin.info/www/bejne/jJqYfISLj7h-oa8

[@G.MOBAStudios]

Bro, how has nobody detailed it like this before? Super appreciate this.

[@taoprompts]

You're welcome! I'll make one for MJ version 6 when it's possible.

[@sammig0203]

I complete agree. The prefer option set is literally life changing!

[@ingridgwladys2827]

Im so happy to have stumble upon this video. Thanks a million. You're the very best 🎉

[@taoprompts]

Thanks, I'm happy to help! Also if you want an updated version of how to get different facial expressions using the character consistency feature, I made an updated tutorial here that covers everything you need to know from different camera angles, expressions, and performing different actions: kzbin.info/www/bejne/jJqYfISLj7h-oa8

[@BigNewsNorman]

Truly mindblowing how well detailed this tutorial is. Absolutely needed this.

[@taoprompts]

Thank you! I'll make a new one when Midjourney version 6 comes out.

[@BigNewsNorman]

@@taoprompts much appreciated man truly!

[@stevenroberts9135]

Yeah and in 6 minutes too! Brilliant.

[@jeremycointin1996]

Great video. consistent characters is a must for us story tellers. I will probably watch this video 5 times.

[@taoprompts]

Thank you, I'll make one for midjourney version 6 soon.

[@DavidFaxon]

This is awesome stuff. I'm a filmmaker just getting into ai filmmaking and I need consistent characters. Is this still the best way to do it with midjourney 6?@@taoprompts

[@XTheBeachIsMyHappyPlaceX]

I had to watch 600 useless videos just to find you explain it so well in a few minutes. Subscribed.

[@taoprompts]

Glad this tutorial helped, and thanks for subbing 🙏

[@freaketonft]

Your way of explaining things is fire! Keep doing it!

[@taoprompts]

Thank you, I'm glad you like these vids, I'll keep them coming!

[@mrLEO.y]

how do you know this!!?? that was one of the most advanced and useful videos I've ever seen about midjourney, thanks a lot !!!

[@taoprompts]

I learn a lot from other people on youtube and twitter. And then a lot of testing! Glad you liked it.

[@theseanwardshow]

My technique for getting bolder, more interesting expressions has been to generate the character with the expression first, then use cref to change some but not all of the features.

[@taoprompts]

You mean you use cref on the eyes or nose specifically? That's a cool idea, I'll have to try that out myself.

[@krleq86]

Unbelievable! I've been struggling to do just this ever since I discovered Midjourney a few months ago. I was doing most of the steps you mentioned, but in a very clumsy order that took hours to yield decent results. Thank you so much!

[@taoprompts]

I think Midjourney is planning to release a "consistent characters" feature soon, but for now this is a good way to get reliable results!

[@nataliesikes4354]

Best tutorial I"ve seen on characters in Midjourney!

[@taoprompts]

Thanks Natalie, Midjourney is working on plans to improve on character consistency.

[@isaacverse]

So much value in just 6 minutes! Keep up the great work, man 🤙🏻

[@taoprompts]

Thanks, I really appreciate that!

[@LoneBagels]

Awesome video man! Sooner or later, this channel will blow up!

[@taoprompts]

Thank you! I appreciate it. I've got plenty more videos planned.

[@thewebstylist]

Just circled back to this amazing Tao MJ tut! Amazing!!

[@taoprompts]

I'm planning on more consistent character stuff soon!

[@thewebstylist]

@@taoprompts GRT to hear mang!!

[@cheyennepesko10]

Very helpful! Thank you. I also purchased your Cinematic & Realistic Prompts eBook. Now I'm getting consistent, moody results. Much appreciated!

[@taoprompts]

That's awesome! Glad to hear you're getting the images you want now.

[@shliggyb]

Incredibly resourceful

[@luluraschi.studio]

Best tutorial on how to create consistency for character/models! Thank you so much! saved to favorites :D

[@taoprompts]

Thank you! Here's a couple twitter threads I learned from, maybe they could help you also: twitter.com/chaseleantj/status/1693246015124713634 twitter.com/chaseleantj/status/1707393560117686472 twitter.com/chaseleantj/status/1700499968690426006

[@luluraschi.studio]

Thanks a lot! Very helpful

[@samnectar]

This is an absolute masterclass in how to perfectly do a tutorial. This information is pure gold, and delivered with the utmost clarity. Subbed!

[@taoprompts]

Thank you & thanks for subbing! I hope this helps you with your Midjourney images.

[@BoxingTayo]

❤ from the Philippines. Wow..fast and precise.

[@taoprompts]

Hello to the Philipines! I hope this was helpful.

[@reidmoto]

Awesome tutorial! I've been searching for a way to generate consistent characters for months, but the results were always inconsistent. Your method creates images that all look like the same character in different settings and facial expressions. Thank you!

[@taoprompts]

Thanks man. Good to know this worked for you. I'm planning to test out some other possibilities for Consistent characters in Midjourney V6.

[@thewebstylist]

Fantastic vid and pro tips as always bro!

[@taoprompts]

Thanks for the support man!

[@jeremyclark4491]

I've watched many guides on creating consistent characters and yours is the most concise and clear that I've seen yet. Great job! What is your strategy for having the posed where they're *not* looking at the camera? For example, a candid side profile shot where they're in a grocery store picking out something off the shelf, or at a diner having a conversation with another person? I've had a hard time inpainting my character without them turning their head (or eyes) towards the camera because all of my reference photos are portrait style looking towards the camera.

[@taoprompts]

I'm glad you liked this guide! Different angles are a little trickier. You need reference images that have multiple viewpoints to inject them into new images. One potential way is to generate the initial split screen image (2 images in 1). Then inpaint one of the square images with a prompt that asks for a different angle like "side profile". This might take a few tries to get a decent looking image. Then if you have a set of "side profile" images you can try using them as references. There's not really a foolproof way of doing this yet though. I'll make an update video when there is.

[@TheHistoriansDen]

I’m big on Meta platforms like Facebook. This is exactly what I was looking for. I wish I could send you super thanks or pay you for this tutorial! Massive thanks again!

[@taoprompts]

I'm happy to help, and thanks for supporting! Your comment helps this tutorial already.

[@Andyg471984]

Thanks so much for explaining it like this!

[@taoprompts]

There's a new consistent characters feature out in Midjourney version 6, I'm testing it out now and I'll release a video on it soon!

[@HumorMeSomeMore]

Any guidance is appreciated with this: I did exactly as far as the expression, but it also changed the face. UPDATE: After watching it a couple of times, I think I got it. I appreciate the time spent explaining the process behind this Tao!

[@taoprompts]

Glad to hear you got it working, it does take a couple of tries to get it right, but it should work if you follow the steps.

[@bdwedgeofanimotion4106]

very nice, will test this out, hopefully someone will be able to create videos with consistent characters

[@taoprompts]

Thanks. Videos with consistent characters would be really cool, that would be super popular.

[@gintsg2053]

Very good explanation and examples. Thanks.

[@taoprompts]

You're welcome! I hope this was helpful

[@giri.goyo_yt]

Excellent! I love how you combined several methods to arrive at your results. Works well. Then I threw one into the free Krea enhancer... WOW. Thanks again, Tao!

[@taoprompts]

Thanks Giri! I did get access to Krea but haven't gotten around to trying it out yet. From what I've seen though it's enhancer is really impressive.

[@rahmabenkhelifa8822]

Could you please tell me how krea helped ? I'm interested

[@goodie2shoes]

you are a wizard. And you have a great way of explaining!

[@taoprompts]

Thanks! I hope this helped with your character designs.

[@janisliepins]

how can i doo it if i have face generated in different ai software and have 1 photo. ?@@taoprompts

[@taoprompts]

If you don't see the option to customize prompts for the vary region inpainting. Make sure you have remix mode turned on. Type "/settings" in to the discord direct message. Then select the remix mode feature

[@Mr.Gahramanov]

Thank you very much for the explanation.so if there are two men in 1 picture, is it possible to change them? for example a man and a woman. For example Emily and Jack. Is it possible to show these two men side by side in different environments?

[@taoprompts]

I will test out if we can do that with the new --cref character consistency feature this week@@Mr.Gahramanov

[@Mr.Gahramanov]

@@taoprompts I would be very happy if you make a video about the cref command :)

[@taoprompts]

I uploaded an update video for the cref command here! kzbin.info/www/bejne/jJqYfISLj7h-oa8@@Mr.Gahramanov

[@edwassermann8368]

very good tutorial. very useful. thanks very much

[@taoprompts]

Glad this was helpful!

[@Juliamawuli]

Thank u alooooooooooooooot !!! Very detailed explanation , exactly what I was searching for ❤

[@taoprompts]

You're very welcome! I will have more content about consistent characters soon 🙏

[@jesseroberts6127]

Nice job on steps and clarity. Other videos I have watched rush through the steps or skip steps. Good work!

[@taoprompts]

Thanks, I'll try to make more videos that are easy to understand like this.

[@DesignWho]

This is wat you call good content.

[@taoprompts]

Thank you!

[@Pusaune112]

I'm always happy when you drop a new video. The informationsare always extremely helpful. You deserve much more attention. If you made more videos, you would definitely grow quickly. Keep it up

[@taoprompts]

Thanks for the support! I'm glad to help. I'll keep making them and try to improve my video editing skills.

[@noamsham2555]

love your content thanx for the teaching. keep it up

[@taoprompts]

Thanks man, these videos are always fun to make. I'm happy you enjoyed.

[@themunki]

This is a breakthrough! 🎉 Instant subscribe 🙏🏼

[@taoprompts]

Thanks for subbing!

[@arturoarvizusart6127]

Excellent video Tao!...super helpful, thanks 👍

[@taoprompts]

Thank you! Always happy to help.

[@Cherie0019]

I love this tutorial. Straight to the point ❤

[@taoprompts]

Thank you! Glad to help.

[@nguyenngocthanhphong1726]

Wow. Big thanks to you! It is so helpful.

[@taoprompts]

You're welcome! Glad to hear this helped.

[@nadiawiesbaden]

Amazing work. Thank you so much🥰

[@taoprompts]

Thanks Nadia, I hope this was helpful!

[@missdiana9222]

WOW 🤯 amazing! Thank you for sharing!!

[@taoprompts]

Thank you!

[@sushmavenkat6552]

Bro, thanks for information. It’s really helpful 👍🏻👍🏻

[@taoprompts]

You're welcome man, I'm glad to help.

[@maelyorn1840]

This is so dope, subscribed

[@taoprompts]

Thank you & welcome aboard!

[@HilalGungorr]

You’re amazing!!!! Your videos are great, you have a great explanation, thank you for sharing these tutorials!

[@taoprompts]

Thanks I really appreciate that! I hope these tutorials helped you out.

[@wrendmc]

Thanks for this!!

[@SvenPiper]

This is very helpful, thx for uploading.

[@taoprompts]

Glad to be of service!

[@aiqinggirl]

great solution!

[@GreenbellyStudios]

Brilliant Ty!

[@MglOhara2099]

i like your channel coz explaining straight to the point

[@taoprompts]

Thank you! I'll try to keep making videos like this.

[@corbin0dallas]

Great detailed instructional video. Thank you! Worth checking out also prettyface ai for consistent characters. Less options compared to MidJourney though

[@taoprompts]

I haven't heard about that ai yet, I will check it out though.

[@aras_mileska]

very very big thank you! 🙏🙏🙏

[@taoprompts]

You're welcome! I'm testing out new Midjourney consistent characters feature and I'll make a new video on that soon.

[@aras_mileska]

@@taopromptsThank you, I look forward to it!

[@nuuttikataja666]

Thanks man!🙂

[@taoprompts]

You're welcome!

[@MordredsRise]

...and the Oscar goes to: YOU! Best Video, thank you 💯👍

[@taoprompts]

Thanks! I'll take the Oscar, glad to hear this was helpful.

[@jalilmeli7027]

good vid, learnt something new

[@taoprompts]

Thanks!

[@ChadRocky]

damn love your channel

[@taoprompts]

Thanks, I appreciate that 👍

[@Zeta_Reticulum]

Thank you, thank you thank you thank you thank you 😍😍

[@Zeta_Reticulum]

subscribed 👍 you saved days of my life :)

[@taoprompts]

Thanks for subbing!

[@Wimpieseigen]

Exellent. Thanks very helpful

[@taoprompts]

Happy to help!

[@ThisIsRabbitHole]

You sir, are a lifesaver. Thanks! --lifesaver

[@taoprompts]

Happy to help!

[@TheBlessingReport]

Great video

[@ChristianHohlfeld]

nice video. well done!

[@taoprompts]

Thank you!

[@elianashapira3310]

This is amazing! Thank you so much! Any update on how to best implement this prompt together with --cref?

[@taoprompts]

Thank you! For getting different expressions with --cref, all you have do is ask for the expression in the prompt "photo of a {description} woman/man smiling --cref image_url"

[@Beanskiiii]

For those of you using V6 alpha you can’t use Vary Region. You have to use 5.2 or below for the Vary Region button to show up

[@taoprompts]

Vary region has been released now for version 6!

[@robertbradshaw1880]

you are the man!!!!!

[@taoprompts]

Thanks Robert!

[@MoJamal-e8q]

epic share. as always. thaaanks

[@taoprompts]

Thank you!

[@dlcanegro753]

awesome bro

[@fetzcou]

HELPFUL!

[@DoctorFashion]

I realize some of these are not an option in Midjourney 6 so I have to switch in order to do this ..

[@taoprompts]

Midjourney V6 is still limited, this guide was made with 5.2

[@SouthernMelle]

Can you advise me on how to keep the clothing consistent from one photo to the next of the same character/pic? Thx!

[@taoprompts]

To keep similar clothing, when you are inserting your reference images, crop out the entire figure silhouette instead of just the face, and then just follow the same process of copying in the reference links for inpainting. Make sure in the prompt you ask for the specific type of clothing like, "wearing a white dress shirt".

[@raphaeldante8277]

Great vid

[@taoprompts]

Thanks!

[@daolaoongdao5226]

Thanks u alot

[@cohanitay5640]

Omg a true tech bro

[@taoprompts]

100% tech bro here.

[@elegost2570]

Thanks for this explanation, super clear and to the point. Is there a similar technique to keep the composition of a character (pose) and adjust other things like background and/or face?

[@taoprompts]

In that case I would use the vary region tool, and specifically select the background to change it. Or specifically select the face to change it.

[@elegost2570]

How the f didn't I think about that before. Thanks man. @@taoprompts

[@finessedancewear7402]

You are awesome

[@taoprompts]

Thanks I appreciate that!

[@gumvue.studio]

thanks,

[@lilillllii246]

Thanks as always. Is it better to fix the seed to keep the routine character or Laura? or midjourney

[@taoprompts]

I don't find fixing the seed that helpful. Do you mean LoRA? That's for stable diffusion, I think it can do similar things as Midjourney although it requires some extra set up. If you're already familiar with Midjourney this way is easier.

[@revko1nyc]

Great video, I wonder if this would work for a design that has a pattern I want, but its some what obscured. I am trying to make a repeat panel out the pattern.

[@taoprompts]

Hey, one tool you can use for a tiled pattern is www.pycheung.com/checker/ All you have to do is upload an image of a tiled pattern and it will repeat it for you.

[@tipfertilizer4595]

cannot cry

[@taoprompts]

Extreme expressions are hard to get with character consistency.

[@elisarondon-muller6650]

GREAT video, is there a way to create character in like children movie animation and you them in different setting. Thinking about something one could use writing a childrens book

[@taoprompts]

If you haven't seen already, there's a new consistent character feature in Midjourney that makes creating characters way easier. I have a video guide here that shows how to use it with illustration/cartoon style characters: kzbin.info/www/bejne/jJqYfISLj7h-oa8

[@O-ZONE-b8i]

A very creative technique. May I ask, now that the character consistency feature has been released in MJ v6 a few days ago, is the method in this video still necessary to further maintain character consistency?

[@taoprompts]

Hey, the character consistency feature lets you change up expressions and camera angles while maintaining the same character, I've uploaded an updated guide here: kzbin.info/www/bejne/jJqYfISLj7h-oa8

[@soytamaravargas8439]

Great video! Thank you! Although I haven't been able to generate the character because when I try to use the prefer option set and put vary region, it gives me this error: "SyntaxError: Unexpected token 'I', "Internal S"... is not valid JSON"

[@taoprompts]

Hey, Midjourney made some changes to the vary region feature and it looks like /prefer_option_set doesn't work the same way. A lot of people are having this issue. A temporary work around is to just use the image links directly after --cref. so "your prompt --cref image_url1 image_url2"

[@Pusaune112]

1:45 the Subtle option would be better at this point or am i wrong ? The subtle option create more similar images than the strong opton

[@taoprompts]

The thing with the subtle option is that the pose & angle of the subject's head is almost exactly the same. I found the reference photos are more versatile with bit more variety in the angle of the head in the photo.

[@NanciFranceVaz_artist]

Hi Tao, excellent tutorial. Is this the same for v6? I am trying to create the same character I generated for a few fine art portrait paintings I want to do as a series.

[@taoprompts]

I haven't tested it in V6 yet, but it should be the same. They have released the inpainting feature for V6 now.

[@alessandroscali]

Hey Tao, thanks for this incredible tutorial, much appreciated! Listen, I have created a character with Fooocus. Is it possible to upload one of more images created with Foocus into Midjourney to achieve your same results?

[@taoprompts]

Yes you can definitely do this with your own images. However, Midjourney typically gives the best results when using images generated directly from Midjourney.

[@alessandroscali]

thanks for your feedback!@@taoprompts

[@olucreativeteam2339]

Great tutorial, does this work with version 6.0

[@taoprompts]

Right now V6 is still limited, I don't think they have introduced all the features into version 6 yet, this guide was made with 5.2 which should work fine.

[@hedmansimon]

Amazing video man! Just a quick Q. im trying to get the "split into 2 images but I'm just getting one images instead of two?

[@taoprompts]

It should generate a single image, that has two photos inside of that image. Kind of like a split screen video.

[@michaelkrogers]

i have a question! say i want to ultimately make a comic book, do i design these characters from the beginning as a comic book style, OR develop them this way and reshape the ultimate images in the drawing style i hope for it to end up in? what would you say?

[@taoprompts]

I would say to start from the beginning with a comic book style if you want to most consistency in the characters generated.

[@willwilliams145]

How do I get the varied region impainting tool? I don't see it in the latest version of Midjourney.

[@taoprompts]

It hasn't been released for the newest version yet, it should be out this week or the next.

[@EdgarAntonioHez]

Would be possible to keep consistency with objects like product images ?

[@taoprompts]

I'm sure you'd be able to use reference images of objects too, using the same method as in this video.

[@fearbabyriffs]

Nice. Do you prefer this to inswapper? I’m a beginner so not sure which is superior. Does this work when to be replaced character has tons on makeup on their face?

[@taoprompts]

Last time I checked inswapper will copy the face over exactly, I don't think you can change the expressions. It's fairly easy to use though so there's no harm in trying. Here's an old tutorial I had on the inswapper: kzbin.info/www/bejne/hqexXqGeh7KXo5I This method should work with makeup on the face.

[@jonyslva]

Really love your tutorials! An this one is mindblowing. I want ask something. Can I use the /prefer feature to also make reference to a specific art style? For example, if I upload 20 images of my art style using /prefer, later in prompt i type “in the art of -jonny” can it work? Or do i need to express it differently?

[@taoprompts]

Thanks man, Yeah, you can use /prefer_option_set to save any settings you want including specific image references, so it should work for art styles also. Although instead of using a prompt like "in the art of -jonny" you can just describe the painting itself like "a painting of a tree --jonny"

[@jonyslva]

@@taoprompts great!! And do i prompt it like the example in this video - -artstyle? Or do i need to express it differently?

[@taoprompts]

yup that should work, as long as you use the same parameter name you save it as using prefer option set.@@jonyslva

[@micheldemolder]

Awesome tut! Can you please close the door next time before recording? Thanks :)

[@taoprompts]

Do you mean the Audio was bad? I have made some changes to improve audio in my newer videos.

[@sylviethomas8625]

Thank you for this. The "split into 2 images" doesn't work for me though.

[@taoprompts]

Here's the prompt I have. It's still working for me in version 6. "portrait photo of a woman, dark red hair and green eyes, pale skin, white shirt, outdoors, natural lighting, kodak portra 160, split into 2 images, shot from multiple angles --ar 2:1 --v 6.0 --s 50"

[@juliataits]

Hi Tao, thank you so much for your videos. For some reason when i try enject the caracter into the image (with vary region) it gives me the error. Unrecognized parameter(s): `--su`, `photo`. Do you have an idea what it could be? The --cref command + prefer option set works fine

[@taoprompts]

It looks like there are some issues with cref and the vary region box at the moment. Maybe they are changing some things in the interface. 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

[@juliataits]

@@taoprompts that’s what I thought! Thank you 🙏🏻

[@sevenoranges]

How can I use my own portrait photograph as the reference image (rather than have MJ generate one from a prompt) please? Can I use a set of five difference self portraits as an option set for reference ? Thanks.

[@taoprompts]

You can try, but Midjourney works best with images generated inside Midjourney. I have a video here that shows how to generate Ai images and videos of yourself by training a lora model: kzbin.info/www/bejne/boane3iIjdGDo8k

[@cheungtiger1088]

Is there a way to achieve the multi - character consistent in on scene?It will be very bursting！

[@taoprompts]

I think you should be able to apply this method to images with multiple people in them. I will test it out in version 6 and make a video on it.

[@linj8794]

what is the max # of photos "prefer option set" would allow

[@taoprompts]

I'm not sure, there may not be one. But I think after you upload 5 images or so, it will be hard to see the effect from each one.

[@felixt1162]

Thanks for the tutorial, I got problems though with adding the links to the /prefer option set, even though I copied the links through midjourney and also tried from my tabs in browser, it gives me the error invalid link! when I test my /prefer option set. Does anybody have an idea how to fix it?

[@taoprompts]

Hey, if you're using discord from your browser. 1. Click on an image to zoom and inspect. This is the one where the background turns black and you just see the image. 2. Then underneath the image click on "open in browser" 3. Then right click and "copy image address" This link should work.

[@felixt1162]

Hey, thx for the reply, I am using it on the desktop app but did it exactly that way and it just doesn't work. After I pasted like 4 urls and some description in my /prefer option set and then want to use the set it will give me the error message invalid link and the link also doesn't work which it shows me in the error message but my original link does work. It's very confusing. I saw on reddit that people had a while ago the same issue but no fix... @@taoprompts

[@jjlee9598]

Very useful..!! Have you successfully made a full body shot image with this character? It is very hard for me to create a full body shot..

[@taoprompts]

Hey, to get a full body shot you can start with a head/upper torso shot, and then use the pan down feature or custom zoom feature to add additional context to the image.

[@jjlee9598]

@@taoprompts That makes sense. Only problem is version 6 does not provide panning at this moment. I guess this will be updated shortly.

[@taoprompts]

Yeah, they mentioned that should be updated in the next few weeks.@@jjlee9598

[@jonahsanchez9371]

How can you edit the prompt after selecting the region you want to change with the vary section option? It only shows submit for me

[@taoprompts]

You have to turn on "remix mode". Just type /settings into discord and a menu should pop up. The click the remix mode button and make sure it's green