How to Prompt for Multiple Subjects in Midjourney V6! - Prompt Guide

[@taoprompts]

If you want to learn how to create MULTIPLE consistent characters in the same scene, watch this video tutorial I made here 👉👉 kzbin.info/www/bejne/pGW4eqyYbqafpJY

[@deathshead357]

Hi there. I wonder if you could help me out? Is there a way I can send you an email?

[@3xxiled]

It still blows my mind that these images are A.I generated. Never ceases to amaze me how I cannot distinguish between generated and real. Absolutely incredible.

[@taoprompts]

For sure, Midjourney version 6 is way better at creating "real" looking photos compared to 5.2.

[@MinkJasonChoreo]

amazing video and editing style. Appreciate the effort

[@taoprompts]

Thanks! I appreciate it

[@doomzy8622]

Spiderman pointing at Batman had me sold.Subscribed🙋‍♂️

[@taoprompts]

Thanks for subbing, we need more spiderman movies.

[@doomzy8622]

@@taoprompts thank you i run a 3k tiktok page under dragonwizart for ai. Trying to build a venomverse of my own but must learn more nasty prompts which i like your style most youtubers are very one dimensional when giving tips have a great day🙏✨

[@Enu_Vibe]

Nice job man. Ethiopian here. Thanks for the shout out.

[@taoprompts]

Thank you, it's a small world man.

[@gunnstark3681]

Fast, Essentiel, nice board tutorial, good job man ❤

[@taoprompts]

Thanks man, I appreciate it!

[@marcusjones1110]

Great video as usual. MidJourney said prompting should be natural for version 6 and you don't have to have super long prompts, but I have been trying to figure how to do it. Your examples of natural language was very helpful. I'm trying to really focus and master Midjourney this year. Please keep the Midjourney videos coming.

[@taoprompts]

I got many more videos planned, prompting is very different now compared to version 5, I'm happy to hear these examples were helpful.

[@marcusjones1110]

@@taoprompts Yeah man...they are very helpful. Keep the videos coming.

[@clade4216]

.. digital art which you may have seen on instagram.. :) made me laugh. love the quality of your vids. keep it up!

[@taoprompts]

Thanks, glad to know you like these vids!

[@NymLily]

Thank you very much Tao, this will be very useful to me. 😊

[@taoprompts]

Glad to know that, I remember you asking about how to get multiple characters doing different things.

[@ChristineBoucher-h1w]

Incredibly helpful!!! Thank you!

[@AiTechRealm]

Nice edit Tao.

[@taoprompts]

Thank you!

[@valcortinas9271]

Great video! Very educational

[@taoprompts]

Thank you!

[@O-ZONE-b8i]

Very useful, it's more convenient than enlarging images to add characters. Midjourney has just released character consistency; can you discuss how to maintain consistency across multiple characters?

[@taoprompts]

That's by far my most requested video, I will test that out this week.

[@O-ZONE-b8i]

@@taopromptsAdditionally, may I ask how to arrange multiple subjects in a scene? For example, a girl is talking to a boy downstairs from a second-floor balcony, with the girl positioned in the top right corner of the frame and the boy in the bottom left corner. These subjects are not aligned on the same horizontal or vertical line.

[@taoprompts]

That seems tricky to do with a single prompt. You would probably want to use a combination of the pan+zoom features with the inpainting tools. Maybe generate a girl on a balcony. Then pan downwards and in the space below here try to inpaint in a boy.@@O-ZONE-b8i

[@mattw8020]

Been playing around with prompting last couple days, this video is exactly what I was looking for, really succinct and thorough. Thanks! Going off your other video of creating consistent characters, you know of a way to do this with multiple subjects in image (changing their expression/interaction), while keeping same exact background? You touch on it a bit with Batman/Spiderman example. But you know if it's possible to apply the 'prefer option set' guide you mentioned ( --anne example) for multiple subjects in same image? While keeping same background?

[@taoprompts]

You could definetley try having two different prefer_option_set of images for different characters and try to inpaint the two characters separately. I haven't tried this out myself yet, but its a good video idea! I will test it out and see if it works.

[@mattw8020]

Having created a character sheet of a subject's face at different angles and using 5 reference photos of those differing face angles, painting in the prefer option set character to multiple subjects one at a time after each upscale does seem to work. But trying to create different expressions with vary regional of the prefer option set subject is proving more difficult. Will keep putzing around with it, the MJ app is pretty incredible though @@taoprompts

[@taoprompts]

That's great to hear, I was wondering if it would work! Different expressions are hard to do even with one character.@@mattw8020

[@denisrodricks]

Thanks, Appreciate it, good as usual

[@taoprompts]

Thanks for the support!

[@mdbess]

Thank you, this was very helpful

[@taoprompts]

You're welcome! I'm glad you liked this.

[@TheBlessingReport]

great video

[@taoprompts]

Thanks for watching!

[@tomasburian6550]

I'm trying to picture a scene with 3 characters, although this vid helped a lot, MJ is still very clueless about how the setup should work. I'll try to adjust my demands to the AI's limitation. Still it has made a great progress lately.

[@thewebstylist]

You are the prompt ma$ter bro!

[@taoprompts]

Thank you! I hope you liked these.

[@wenstoria]

thanks tao

[@EricWang-j5z]

Great I love it

[@taoprompts]

Thank you, I hope these prompt ideas help.

[@ptardo22]

Tao I 've trying to get interaction between characters in the last months. What you've got i'ts a mine gold. I guess what worked on bat & spidey it's that you didn't need to add tomany details on the character, so midjourney primary focused on the interaction beetween characters. Did you try to do it backwards? Focusing on the interaction, and then adding details to the characters?

[@taoprompts]

Thanks, good to know this was helpful. I find that Midjourney still struggles a lot with character interactions, there's just not that much you can do yet. I usually try the interactions first and then add in details.

[@sarahashemi6521]

Thank you very much❤. I like your tutorials. They are great. Could you please make a tutorial to show how can i repeat a sme background in multiple images? Im working on an anime style series. In many scenes, the character is in the same room in his house. Im new in midjourney, and i don't know how can i use s same background. Thank you ❤

[@taoprompts]

Try using the Midjourney editor for inpainting. Or create characters and backgrounds separately and add them together

[@Pandorocha]

Thank you very much.

[@taoprompts]

You're welcome, I hope this helped.

[@motivationzumhoren2725]

Thanks a lot for ur great content.❤ Can u help me with the realistic texture of drenched clothes. I creat some rainy scene nd my subject looks photorealistic but hair and clothes are dry. Biggest challenge is also if u use diffrent fabrics for a rain scene or fight scene of warriors in the muddy landscape.

[@taoprompts]

I tried a prompt like this "a man in the rain, his hair is wet. He is wearing a wet white collared shirt --style raw --v 6.0" and it seems to work. It could be that the specific fabric or clothing your using is hard for Midjourney to handle. Maybe try using raw style and describing how the clothing is wet and covered with water.

[@motivationzumhoren2725]

@@taoprompts 😉🙏🏻thank u for testing. I will try to describe more specific.

[@maim3170]

when i create a scene with the character, the character is always too big for the scene. How do i place the character in the back scene as sitting or looking aside?

[@taoprompts]

It does like to make characters & subjects in the forefront of the images. You could try creating an image with a character you like, then use the zoom + pan features to pull the camera back so you're character looks smaller relative to the scene.

[@puja1985]

I think the inpainting feature in v6 will significantly change the game. Finally, Midjourney is stepping up their game.

[@taoprompts]

Yup, I'm looking forward to the release, hopefully within 2 weeks or so.

[@emm1924]

Merci super tuto 🙂

[@taoprompts]

Heureux de pouvoir vous aider!

[@parodyverse]

Thanks for this video. What would a prompt look like if I was trying to generate a classical band playing in a mansion? Every time I generate the image the faces are mushy.

[@taoprompts]

The more and more you zoom out and get smaller faces the more messy the Ai will be. One thing that helped me was using the "creative upscale" feature which seems to improve facial detail.s

[@MinkJasonChoreo]

With each step you mentioned is it a fresh 'imagine' or are all of these triggered from changing the prompts after selecting 'vary strong or subtle' ?

[@taoprompts]

Those are all fresh imagine images

[@박진현-x8w]

Thanks for very helpful contents I have a question. I want to apply --cref to multiple characters with one prompt each, how can I do it?

[@taoprompts]

You could try to generate an image with 2 people in it. Then use the vary region inpainting tool on each face separately with --cref of a different character for each one. I will try generating multiple consistent characters myself also.

[@pilot926]

good vdo ,how to create consistency character on v6?

[@taoprompts]

Now that V6 has released inpainting and Pan features you can use this tutorial here to get consistent characters: kzbin.info/www/bejne/iIHbdpptnteqd6M

[@Wimpieseigen]

Thanks

[@BibiTheLinkBuilder]

I'm trying to combine movie characters and it keeps mixing them up intro weird blends

[@taoprompts]

It does depend on which specific characters you are trying to combine, batman and spiderman seem to get mixed up pretty often. Using a wide aspect ratio usually helps give them more space on their own to not get mixed up, but a lot of the times you need to reroll the prompts to get images you're happy with.

[@BibiTheLinkBuilder]

@@taoprompts ahh this is good to know! I guess its similar to doing an actual photoshoot and taking loads of shots to get a handful great ones :)

[@taoprompts]

Yeah exactly, sometimes it takes a couple tries and variations to get something workable.@@BibiTheLinkBuilder

[@amumtomum]

How can I make consistent multiple characters in one image?

[@taoprompts]

Hey, here's a video tutorial I made on getting multiple consistent characters in one image, kzbin.info/www/bejne/pGW4eqyYbqafpJY

[@catherinemok2481]

I am a midjourney beginner with no graphic experience at all. What can I download or read to make myself better?

[@taoprompts]

Hey, Midjourney is still a really new tool. That would depend on what you want to use Midjourney for. The Midjourney discord channel called "prompt-faqs" is a pretty decent place to get advice and answers a lot of common questions. If you want pdf prompt guides I have a few here: taoprompts.gumroad.com/ Chase Lean or Nick St Pierre have a twitter accounts and posts good Midjourney Tips from time to time. Otherwise, there's a lot of great Midjourney youtube channels like, Future Tech Pilot, Cyberjungle, Tokenized AI.

[@ShruthiJoshi-t7c]

Hi can you please help me with writing prompts for multiple Indian mythological characters?

[@taoprompts]

What happens if you try replacing the characters in my prompts like batman and joker, with specific Indian mythological characters?

[@Flux_Nomad]

Lol I came here wanting to do a silly Kaiju Battle between 2 different monsters, but of course it keeps blending or sharing details or making 2 of the same monster. I experimented with just a man and woman like this video... okay works. Tried replacing the man with Robot and then both take on the attributes of a Robot; woman is kind of Cyborg. Interestingly, I replaced the Man with Male Skeleton and it works in half of the pictures lol.

[@taoprompts]

Midjourney does blend subjects together depending on their characteristics. Sometimes it takes a bunch of different generations to get one that looks decent. From there, I'll create variations and use the inpainting tool to get an image that works.

[@catherinemok2481]

The picture profiles I upload does not look like him/her using Inswap. Please advise.

[@taoprompts]

Do you mean you are trying to swap in your own pictures with insight face swap? That tool will swap the eyes, nose, lips area from your picture into the image, but the hairstyle and other head region doesn't get swapped. It is tricky to map yourself into an Ai image without photoshop. Lora/Easyphoto/InstantID with stable diffusion work decently, but will take some extra work on your end. You could try using the consistent character feature in Midjourney, but results will be mixed.

[@AIBAXperience]

Nice tips, but easy with 2-3 peoples, but after that it's hell on earth to produce good face for groups or crowds.

[@taoprompts]

It does struggle with faces for multiple people, that's one of the main limitations atm.

[@jl407b]

I will subscribe your channel if you answer my question The question is : Can I able to create all celebrities ai generated photos ?? Like example : Ishowspeed sitting with messi in the coffee shop

[@moochilinibailey4419]

well the hispanic woman prompt is too vaugue because the black woman can be hispanic

[@taoprompts]

In real life yes, the Ai model doesn't understand that much nuance though. For example, an asian person could be German, but if you prompt for a "german man" it will always generate a caucasian person.

[@DGFilmsNYC]

the black woman on the right could be a dark skinned hispanic woman ijs

[@taoprompts]

That's true, although generally if you prompt for someone hispanic, Midjourney always generates someone with lighter skin tone. If you test it out enough times, you'll see that Midjourney does in fact have trouble generating subjects with certain combinations of ethnicities together.

[@Tyrell_Corp2019]

I appreciate the video but I'm still not keen on spending money with these services. None of the AI out there can get close to what you want even with a reference photo. It takes A LOT of run-throughs. And then that's just one shot. It's very time consuming. Maybe in a year we may get there. But I don't think so.

[@taoprompts]

If you want a really high level of control, you might be able to get that using Stable Diffusion + Controlnet + LoRA, although that will require more setup on your end than using Midjourney.

[@Tyrell_Corp2019]

@@taoprompts Thanks. I’ve seen videos on that. I’m considering it. :)