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

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
let's walk through how to prom for multiple characters in mid Journey step by step the version 6 update improved the flexibility of mid Journey quite a bit and there's a lot more you can do now to generate multiple characters but there's still some limitations we need to talk about and thanks to Nick same Pier for inspiring this video make sure to go and check out his Twitter account Step One is setting the scene we can start with a basic idea of what we're looking for like two people are sitting at a poker table that's simple enough don't be too specific at this point we just need a general idea with as few details as possible it's also helpful to introduce a medium for the image I want a photo so I'll add Shout with Kodak portra 400 film to get a wide image I'm going with a 2 to1 aspect ratio and I also use raw style to make sure that mid Journey follows my prompt as closely as possible so this image is reasonable but I think we could actually be a bit more direct with what we want want I'm looking for a front view and to keep it interesting I'll prompt for a man and a woman sitting together remember at this stage we just want to frame the subjects in the context of the surrounding environment without being too specific with the details once we have the basic structure of the image down we can start to adding more specific details this is a fun part where we get to play around with different ethnicities clothing colors accessories and other details you can think of here's an example of just how fine grain the prompts can be I specifically prompt for a young Ethiopian man wearing a light blue button down with a Rolex and an older Asian woman in a yellow turtleneck and dark blue peaco notice how I asked for the man to be on the left and the woman to be on the right you can tell mid Journey exactly where each subject should be in the image and this actually helps with image Generations also notice how I'm using natural conversational language in the prompt if in mid Journey version 6 avoid using sporadic words like man in blue shirt and instead try using more natural language in your prompts for the best effects mid journe is able to generate all the details we prompted for from the race of our subjects down to the specific clothing details even the Rolex is placed on the man's hand let's try another example we'll ask for an older white man on the left in a black tuxedo and a younger white woman on the right in a white T-shirt with blue polka dots she's also got blonde hair this time we'll ask for some interaction with the man looking at his phone and the woman looking directly at the man the details look correct and the amount of control you get is amazing the clothing is exactly what we asked for even down to the blue color in the polka dot shirt the interactive part is lacking though you'll notice that the woman doesn't look directly at the man in these photos like I asked for something worth noting is that mid journey is still kind of limited by their training data sets for example it's hard to get certain ethnicities of people to be in the same image here I prompted for a young Asian man and an older Irish woman even though the clothing details match my prompt it ends up generating two Asian people this was supposed to be an image of a black man and a Hispanic woman and the AI generates an image of two black people this is something you'll have to play around with to know what types of combinations are possible at the moment let's test out Interactive movement between our characters a bit more here I asked for the man's arm to be around the woman's shoulder M Journey makes them lean against each other but it was hard to get the specific arm around shoulder motion that I wanted for the most part in some instances it gets mixed up I prompted for the man to have his arm around the woman but funny enough moury does the opposite and puts a woman's arm around the man here's an example where I prompted for them to be holding hands again we expose some limitations mid Journey still has it can generate really fine grain details in the appearance of the subjects but interactions between subjects are still a struggle especially as you add more and more details if you prompt for them to do separate things like hoding items that can work slightly better this prompt asked for the man to be drinking a cocktail and the woman to look at her phone and both actions do appear to be in the image make sure you also experiment with the different locations and B backgrounds for your subjects whether that's going for a jog in Arizona with red rocks on the side of the road or jogging next to a Scenic Lake the background you use can add a lot to the atmosphere of the image Batman and Spider-Man are sitting at a bar Batman is on the left wearing a khaki peacat Spider-Man is on the right wearing a gray hoodie Spider-Man is pointing his finger at Batman take note of the details here Batman Spider-Man at a bar Batman in a khaki peo Spider-Man in a gray hoodie Spider-Man points finger at Batman that's enough arguing time to shake hands and make up even though there's some limitations in the types of interactions that are still possible it was fun to create scenes for some of my favorite characters these guys look like they had a la day at work a concept you should know is using callbacks in your prompt this just means referring back to your subjects as you add more and more details a man and a woman sitting side by side at a circular table on the left is an older white man with a gray mustache wearing a light blue button- down shirt on the right is a younger white woman in a black cocktail dress they're in a sunny outdoor patio with vines on the walls the older man is eating a garden salad and the younger man is eating a turkey sandwich notice how I'm referring back to the man and woman multiple times as I add more details for each one of them individually callbacks are really important to use so that mid Journey knows exactly which subject you're adding details for it's easy to switch up the medium that the image is generated in usually it works better if you add it to the start of the prompt we can use oil paints which have Rich B colors and a wide range of textures watercolors which are more transparent with less saturated colors digital art that you may have seen on Instagram pencil drawings and other styles you might think of if you want to add more characters just follow the same formula remember that it's really important to specify in your prompt where each subject is located relative to each other I want the Hulk on the left in a red track suit Black Widow in the middle in a blue shirt and Tony Stark on the right in a black leather jacket it always looks like superheroes are having a bad day when you generate them in mid Journey let's review what we've learned step one setting the scene use a basic description of the general image you want to generate without adding in too many details step two adding the details explore different clothing colors textures ethnicities age mid Journey can generate really specific prompts especially for clothing and colors now step three character interactions this is still very limited in mid journey I try to make Batman and Spider-Man punch each other but that just doesn't work step four experiment with different locations and backgrounds for your subjects and step five try out different mediums for your images use different styles like photography or oil paintings for different visual effects now I know people want to know about generating multiple consistent characters we'll see if that's possible when mid Journey releases its imp painting feature so make sure to subscribe for that also if you want more ideas for your prompts go and check out this prompt guide I made [Music] here
Info
Channel: Tao Prompts
Views: 6,158
Rating: undefined out of 5
Keywords:
Id: F53ti2sD7KI
Channel Id: undefined
Length: 8min 1sec (481 seconds)
Published: Mon Jan 22 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.