Midjourney Character Consistency For Your Stories GUARANTEED! (NO SEED NEEDED)

many of you have asked me how I maintain the consistency of my character throughout my whole animated stories so today I will be covering this in this tutorial so before we start I wanted to briefly mention this first method called the clarinet's puppet methods which basically was the first method I used in the beginning and it does still work well I'm not going to go too much into detail with this method but as you can see here is the template to follow first we have the image link of our character followed by setting is so here just add the location then we have name is so here just add your character's name and briefly physically describe him or her at last we have set and contains name which is the name of our character doing activity which is where you add the action you want your character to do then add the atmosphere filler Etc and we have an example right underneath here I'll put the link in the comments section if you want to have a look we won't need it for this tutorial so the first step of our method consists in first generating an avatar description with chat GPT so ideally by now your story is ready to go and you have all your characters listed let's say that we wanted to make a story about a dancer named Jason what we are going to do is to ask chat GPT to physically describe a male dancer for my upcoming story his name is Jason it will give you quite some text but we don't need this huge text so we're going to ask give me a shorter version here is what it returned to us Jason is a tall and lean male dancer with a chiseled physique sharp features piercing blue eyes and dark hair styled in a slick Manner and this is only what we need so we're going to copy just that first part and head over to our mid-journey account once here just paste it as a prompt but we're going to add the word portrait of and this is what it generated for us I really like number two so I upscaled it and now let's save it click on the picture open in browser right click save images and save it I've already done that so I'm going to skip this time but you would want to click save and then go back now here there is this background and we don't want that because later it sometimes interferes with our images so we will remove it I use canva but any background removal software will do once you import your image to canva click edit image and then background remover I've already done it so if you had to do it just go ahead and download the image make sure that transparent background is ticked quick note sometimes it's best to crop to only the head as sometimes the rest of the body gets incorporated into our generated images you will soon see what I mean but I will leave it as it is here now let's head back to Mid journey and let's import our new image click this plus button upload a file and we're going to choose our picture click open and the picture will be generated now here's where it becomes interesting so the way I approach this in my latest animated story was instead of focusing on my character and trying to fit him in different settings like in the clarinet puppet method I focus first on creating scenes that already include a random character who physically looks like my character doing my desired action then later focused on matching my character face with this random person's body it might be a bit confusing for now so let me show you it will make sense in a moment so let's say that in our script the opening scene of our story introduces our dancer Jason dancing in the club I don't have any specific requirements just that it has to be Jason dancing in a club so I went ahead and wrote this prompt wide full body image then used our chat GPT description lean male dancer with a chiseled physique sharp features piercing blue eyes and dark hair styled in a slick manner sometimes I add the name of my character at the end of the description to teach mid-journey to start associating the name with my character but this time I didn't I then added my desired action which is dancing on the dance floor followed by the location in a club full of people aspect ratio 3x2 so here's the result but as you can see our Json is topless which is not what we want not at this time anyway yeah boy so we can re-roll but this time I decided to add in my prompt wearing fashion clothes still dancing on the dance floor followed by my desired Style if you wish to add your desired style like an artist's name or a specific texture feel free to do so but make sure to integrate it into most of your future prompts to keep the style consistent throughout your story now here are the newly generated images it's similar but in some we have some clothes still not what I'm looking for but I liked an upscaled number one to use it as a live demo later on so I decided to be more precise in my prompt because I didn't want a topless Jason I went ahead and replaced wearing fashion clothes with wearing a black shirt jeans and shoes and as you can see I really like number three this one is already pretty close to our Json so I upscaled it and decided to re-roll just to give us something else to work with so that I can show you the step process with a different picture here I really like number four and it doesn't quite look like our Json so it's perfect for our demo I upscaled it and re-rolled again to give us something else to work with you don't need to re-roll but I like to have multiple options like in many movies many scenes don't make the final cut but at least it's good to have many options to choose from so right here as well I really like number two so let's upscale that one two while all our latest images are being upscaled here's our first generated image we haven't done anything yet and it's already looking good I personally would be happy with this and would go ahead and save it as we normally do open in browser right click and save as now this is where our technique comes into play as you can see the face is quite different from our Json so how do we replace it step one collect our image link click the image right-click copy image address then head over to start a new prompt imagine prompt paste our image link Step 2 collect our Json face link same thing as step 1. click the image of your character right click copy image address then head over to the already started prompt and paste it next to the first link no need for a comma step 3 add the prompt from our original image we're going to copy our prompt right here copy and we're going to paste it next to the two links that we already pasted and hit enter I've done it already and this is what came up so as we can see this is pretty good and quite close to our original image remember I'm not trying to be too specific the main idea is to have Jason dancing in the club also here we have our long links shortened the first link is our reference image and the second link are Jason's face so before we move to a live example here is a quick recap one we asked chatgpt to physically describe our Json we shortened the long version and took only the part describing Jason's physique two we generated Jason's portrait in mint Journey upscale our chosen one and removed its background once done we imported our new image back into mid-journey free we generated an image of our ideal scene by using our chat GPT short description and combined it with an action a location and a style four we saved the images that were already looking accurate in our example just one and if we liked one that wasn't so accurate we upscaled it copy and pasted its Link in a new prompt copied and pasted our character image link next to our first image link followed by our scene prompt this is how our prompt looks like image link One Plus image link 2 equals image 3. okay so now back to our results picture number three is the one that stands out the most for me in terms of how accurate it looks to our Json so I upscaled it quick note to make it easier for us let's save our Jason's face shortened link into a notepad page we're going to call it Jason's face and paste our link now you can see this minor detail on his chin that isn't representative of our Json so I decided to make some Variations by clicking make variations and it was much better we now have four accurate images that represent our Json dancing on the dance floor in the middle of the club now let's quickly do a live example with the other images we upscaled earlier let's use this one same process click the image right click copy image address imagine prompt followed by a reference image link then we're going to copy and paste our json's face link that we saved a moment ago and add next above links our prompt that contains the description settings Etc hit enter as you can see despite our prompt our Json is topless and this is due to not cropping our image to just the face it's a bit special also there is some similarity but it's not exactly a match to what we want so we're going to re-roll it and see what we can come up with different sometimes that will be the process the re-roll button will have to be your best friend so here we will re-roll again as it's still not what we want we want to be accurate with our character but not so accurate with our desired scenery our image is now ready image number four is quite close to our character I really like it so we will upscale this I think that we can work with it while our image is being generated let's start working on the other image we upscale a while back click the image right click copy image address imagine prompt followed by our reference image link copy and paste our json's face link that we saved and let's add next to both links the rest of our prompt hit enter you can already see that we are getting a good representation of our reference image okay now we have our upskilled image ready and it's quite representative of our Json with our original image I would go ahead and save it now coming back to this one you can see that number one stands out and maybe number four too we could re-roll one more time or we could upscale them and separately try them as New Image references we would then paste our Jason's face link image followed by our prompt now let's say that we wanted to work with a new scene our Jason is reading a book in the library for example we would write imagine prompt wide full body image followed by our chat GPT description lean male dancer with a chiseled physique sharp features piercing blue eyes and dark hair styled in a slick manner wearing a black shirt reading a book in the libraries super sleek and our aspect ratio let's see so now this is what we got and you can already see that the face and hair features of number one two and three are similar to our Json you can also see that image 2 has the open shirt and chest visible again which I believe is due to not cropping closer are Jason's face earlier so mid-journey is still giving us that option but we won't use it having said all that image 4 is what I want so we're going to upscale that here we now have our image now let's apply the steps used previously click the image right click copy image address imagine prompt followed by our reference image link copy and paste our json's face link and next above links add the rest of our prompt hit enter now something that I want to add while our image is being generated is that when you start adding and combining your prompt with different views and angle shots like low angle view high angle view medium shot long shot Etc it helps in making your movie Stand Out by diversifying and giving more than one perspective from the traditional normal view if that makes sense also when applying this technique if you need a back shot make sure to remove any description about eyes or face from your prompt it's usually a 50 50 result if you leave this detail finally instead of generating images from mid-journey for your image reference you can use real life images too for example in my latest video I used this doctor strange scene as a reference to create this kind of out of body image and then applied our technique to integrate our character I then made sure to add in my prompt the action I wanted my character to do or experience by writing out of body experience to really help mid-journey understand where I wanted I later added some editing to bring it to life you can also do this if you want a specific location from any images just import it in mid-journey and add the link next to the other links in your prompt so now your prompt would have your first image reference link followed by your character face link and then your location image link now let's go back to our results so here number four stands out quite well but let's give it a re-roll one more time and see if we can get something even better look at this these are really close to our character now because it's mid Journey not every picture will be identical but it will give you at least a 90 to 95 success rate in maintaining your character here we could always re-roll if we wanted to get an even closer match and see what mid-journey can give us or we could apply the step previously mentioned where you would just upscale the closest image to your character and use that new upscaled image link as the new reference image then add our Json image link and our prompt if you have any questions leave them in the comments below if you would like to hire me to work on your story email me and let's see what can be done the next tutorial will be on how to come up with great story ideas that you can combine with chat GPT and I will also have a few more on the way I edit and animate my images so stay tuned for that so yeah thank you for watching and I'll see you in the next video

Channel: Ai Fiction Factory

Views: 9,498

Keywords: Midjourney Character Consistency For Your Stories GUARANTEED NO SEED NEEDED, midjourney character consistency, midjourney, midjourney v4, midjourney ai, midjourney tutorial, midjourney prompts, midjourney scenes, midjourney storytelling, midjourney consistent character, midjourney consistent style, midjourney character design, midjourney multi prompts, consistent character midjourney, consistent character, ai fiction factory, ai art, Chat gpt, midjourney character art, chatgpt

Id: dqcIOOmmlaQ

Length: 15min 14sec (914 seconds)

Published: Sat Feb 18 2023