Cinematic images in Midjourney - Several characters in a scene

foreign we worked with creating a consistent character in a cinematic style working from a script we found a look for our main character svenia and we started placing her in location for scenes from The Script our next stage is to find the look for other characters from the script and then to place them together in scenes placing several consistent characters together is easier said than done right now as a version 4 in the middle of February mid-journey is outstanding at doing a single subject great at placing that subject in a setting but when it comes to placing two subjects together in a setting mid-journey mostly fails the solution to this problem right now is to use Photoshop or other image editing software to get the results you're after my guess is that future versions of my journey will have great functionality for consistent characters and how to combine them in the meantime we tinker and adjust as best we can first we need to create a second character in the first scene in The Script spinya meets her new boss for the first time all the info the script gives us is his name and age Jefferson aged 50. but from the dialogue we get that he's a bit pretentious and narcissistic soon sparking a conflict with Spina so how does he look first we try to generate a photographic portrait using simple prompts until we find a face we can work with the first few attempts doesn't give us the character we're looking for with small adjustments to The Prompt a character pops up that we'd like to explore further we upscale that image and use it to place him in one of our location we use one image prompt link for our character and one for our location from experience we know we need a short text prompt to go with the images to steer Us in the right direction style wise pretty soon we get a handful of images of Jefferson that we can use for later image prompts but the next step is the trickiest how can we place two consistent characters in a scene together first we can test if my journey understands basic film language like the main terms for having two characters in the same shot over the shoulder shot and two shot putting our characters aside for a moment we prompt for over-the-shoulder shots our resulting images show that the basic composition can come out okay but majority Blends elements in an unwanted way things like beards have a tendency run amok let's try to simplify the prompt further before we move on this time we just prompt for a man and a woman having a conversation from the results we can see that the style falls apart so cinematic terms like over-the-shoulder shot actually helps us maintain a cinematic Style our experiments show that the term two-shot isn't helpful instead we can just describe how two characters are positioned in the shot and what their action is as we can see from these tests we need to steer towards our style using image or text prompts or a combination of both the most intuitive thing would be if you could feed my journey image prompts of your two characters and just write something like two people talking but that doesn't work in this version we can get the style to where we want with simple text prompting in addition to the image prompts but your two characters will not come out consistent so we need some workarounds the first method we're going to look at takes advantage of the over-the-shoulder composition in this shot type we often only see the face of one of the characters clearly let's try combining three image prompts with a text prompt one of the reference images shows an over-the-shoulder shot while the two other images show our characters the text prompt tries to set our style and composition this method gives us okay results that we can upscale and do some more tweaking on method number two is to put together a reference image in Photoshop that we can feed back into my journey as an image prompt nothing fancy just the two characters cut and paste it together in a location the important thing is seeing what majority can do with it using that photoshopped reference image together with a prompt gives us promising results but there are signs that the Cinematic style and especially faces and eyes are falling apart we try to add another reference image prompt of spinia's face to hopefully make her look more consistent all in all this method works pretty well but needs a fair amount of Photoshop work on the generated images to fix wonky eyes adjust colors add texture to skin and so on my third option uses remixing to add characters faces first we try to generate a good over the shoulder shot just using a text prompt if we get an image we'd like to use we remix it pressing the V button connected to the image this pops up the remix window where we can adjust the prompt here we add an image prompt of our main character's face the method can give okay results but needs a lot of tinkering to get the initial prompt right with hairstyle clothing colors location and so on we can try to combine method 2 and 3. first we use our photoshopped reference image of the two characters in the location together with a text prompt then we try to remix with an image prompt of our main character's face to strengthen her consistency the results here shows clear signs of falling apart style wise so sticking to the simpler methods give better images to sum up placing two consistent characters together in a majority image right now is complicated there are methods that give okay results but they need a lot of tinkering and adjustments as well as Photoshop work what do you want us to tackle in our upcoming videos different cinematic genres and their unique looks lighting action scenes leave your suggestions in the comments below thank you

Channel: CinemAI

Views: 7,129

Keywords:

Id: lmCnOIQq_5c

Length: 6min 47sec (407 seconds)

Published: Wed Feb 15 2023