Mastering Midjourney v5.1 : A Deep Dive into the New Version! (RAW, Seeds, Tips and Tricks!)

hey everyone today we are taking a deep dive into mid Journey version 5.1 it's a really great update with a lot of imagination but it can also be a bit much so I'm going to show you how you can maybe tame that back a little bit I've also got an argument for V4 of all things plus a trick for Mid Journey's least used feature at least I think it's the least used feature okay let's dive in so version 5.1 is a pretty interesting turning point in mid-journey history it's the first time we've had a numbered update between versions and it's actually two different modes within version 5.1 5.1 is a more stylistically opinionated model one that kind of more resembles the shorter prompting style of version four 5.1 also has raw mode which I think a lot of people are mistaking to be just version five and it's not it's really its own thing we're going to take a look at that in just one second 5.1 is now also default but just in case you wanted to switch things around all you have to do is hit forward slash settings and then you can control any of the previous versions plus 5.1 or raw mode so in order to accurately assess each version I not only wanted to use the same prompt but I wanted to use the same seed I thought that might make a good opportunity just to very briefly go over seeds because I think sometimes there's some confusion about them so each job that the mid-journey bot generates has a seed number to it that seed basically is almost like television static that eventually forms into the image that you have prompted so your seed number is basically that initial pattern of static so you'll never get the same image exactly twice but it'll be relatively in the same ballpark it'll start at least in the same place as an FYI the seed numbers are finite and I believe randomized as well so if you go back to a seed number that you had five months ago it may not be the same seed number as it was before if you go back to the same seed number that you rolled 24 hours ago it might not be the same it's all just sort of luck of the draw in order to find out your seed number and actually as a quick update as well all you have to do is come up and add a reaction to any image use the Envelope Filter and you'll get a message with that seed number so previously due to a glitch in version 5 you could only get the seed number of the initial for grid image apparently as of last month that glitch has now been fixed so you can now take your upscaled images and find out the seed number speaking of which let's go generate an image grab the seed number and take it through the various version so I began by generating in version 5 with kind of a bond inspired prompt it was cinematic shot spy film a French supermodel wearing a black dress seductive expression waiting at a high fashion bar atmospheric lighting with an aspect ratio of 2-1 and this was our image which looks nice it's a little bit on the new Ori side to me with that heavy shadow cast on the side of her face but overall it's a really nice image so taking that same prompt and seed and running it in version 5.1 got us this image which is definitely much more stylistic than our initial V5 output the skin tones are a lot warmer and a lot brighter there is less Shadow on her face the background definitely pops a lot more than it did in the initial V5 image and I do feel that there definitely is more volumetric Lighting in the background as well so now popping over to V5 raw once again same prompt same seed gets us this image where the background mostly stays the same there are some subtle differences in the lights overhead but really the major difference here to me is in the skin tones version 5 gave us very smooth very soft very uniform skin tones whereas in version 5.1 raw there's a little more white peeking through it doesn't look quite as glossy I mean it still looks obviously like a French supermodel like there aren't any gaping pores or wrinkles like like you know normal people have but still it's a French supermodel I'm willing to let it slide um overall though I do think that V 5.1 raw tends to look just a little more on the realistic side and I'm not bagging on the base 5.1 model either I mean I more or less was calling out a Bond film without calling out a Bond film so maybe that kind of glossy Perfection is something that it should have overall version 5.1 is considerably more imaginative and expressive than version five and that's especially noticeable when you run very basic prompts like in this case I ran donut and version 5 gave us a very photographic standard delicious looking donut a 5.1 on the other hand gave us these illustrative kind of playful and vibrant images of donuts so there's nothing wrong with either version It's just good to know that that's kind of where each model is thinking another look at a very basic prompt example this was happy kids playing with toy robots version 5 gave us this which is very a stock photo Sears catalog kind of imagery version 5.1 outputted this which is pretty interesting the fourth image is really the only image that has a standard photographic feel to it even then it's a very heavily stylized with the Forefront being warm and the background being that colder blue depending on your intentions that level of stylization may work for or against you consider the following image which has the very simple prompt of a young boy fishing at a lake on a dock summer golden hour medium shot ar-169 version 5 has a very nice photographic look to it it would look quite at home as a stock image or as a documentary image whereas version 5.1 is a lot more stylized it does look like it's been through Photoshop a few times and actually in one of the outputs definitely has more of an illustrative style to it and that brings us to Raw mode which I think is a really good compromise between the heavy stylization of version 5.1 and the kind of well what now sort of appears to be a little bit on the Bland side of version five so starting with a very simple prompt in version 5 with a cabin in the woods we got this image which I thought was you know pretty nice I definitely that fourth image is probably more what I was looking for because whenever I hear Cabin in the Woods I'm gonna instantly think it's you know a horror movie so 5.1 kind of surprised me by returning these almost Thomas kinkade-esque paintings it was not at all what I was expecting I mean granted it was a very basic prompt and I wasn't spelling out the tone or style that I was looking for but it's interesting that it defaulted into this kind of warm autumnal cozy cabin Vibe but 5.1 raw returned as these which is actually kind of a nice merging between the two it isn't overly kitschy like 5.1 but it isn't quite as dreary and colorless and drab as version five so by adding to this prompt in version 5 raw with cinematic still Cabin in the Woods heavy atmosphere Knight volumetric fog and horror film lighting we end up with these images which I think are really great images particularly that fourth one which is probably what I had in mind when I initially typed The Cabin in the Woods basic prompt um so yeah in my opinion raw mode is the best mode to work with it allows the creativity of version 5.1 without that heavy layer of stylization over it that said raw mode is best suited when you have a very specific image in mind and is probably better served with longer prompts I did want to note I think there's been some additional training in 5.1 as well I ran across this while doing some experiments trying to emulate the style of Charles Peterson Charles Peterson is a photographer who documented the Seattle grunge scene in the early 1990s photographing bands like Nirvana and Soundgarden and many more Charles's photos are filled with chaos and movement motion blur and light blur and everything has kind of a dirty grain to it which is the perfect aesthetic style name dropping him into mid-journey it was pretty clear that mid Journey had not been trained on Charles's images but I began to notice something interesting so I have these guitars behind me they aren't here for decoration I do play I have played for many many years and one thing that mid Journey has never really been great at is fingers and instruments and fingers playing instruments particularly guitars and I think that's something that guitar players always kind of tend to pick up on but since 5.1 I'm starting to see a lot more actual correct fingers things that look like they are really playing the instrument and by no means am I saying it's completely gone away I mean we we have this guy with one two three four five six seven fingers uh on his hand which actually some guitar players would love to have um but for the most part I'm actually starting to get more accurate guitar fingers than I've ever seen before and I did want to point out that as I was continuing on with this Charles Peterson experiment um the place that I actually found the most success weirdly enough and it's not completely in the style but it's closer to the energy of what I was looking for was in mid-journey V4 I think V4 was better at capturing that kind of grainy black and white chaotic look then V5 was I think that we're starting to get to a point where we are at the point now where choosing between version 4 and 5.1 are starting to become aesthetic choices like they all produce really solid results so the question is what are you aiming for and which version is good at providing that thing that you're looking for before we move into the next section I just wanted to let you know that there is a PDF of today's video over over on gumroad it is free but you are welcome to leave a donation if you so choose and I'd like to thank everyone that has left a donation so far it really does mean a lot all right let's hop into the next section so one very noticeable Improvement of version 5.1 is in the sharpness um that's something that I noticed really took a hit with paper quilling type images in version five version four was pretty solid version five yeah things got really soft um so I'm really happy to report that in version 5.1 the sharpness seems to have returned so I think between version 5 version 5.1 and version 5.1 raw we've got a really interesting use case for permutations uh permutations I think are probably the least used feature of mid-journey and probably the most widely misunderstood as well and I get why they're fairly confusing to use they burn a lot of fast hours and I feel that most people never really found a particular use for them if you haven't run across permutations or are still confused as to how it works I did a whole video on it you can check it out it's linked below the new versions really open up a door for trying out permutations so I gave it a shot by taking a cinematic still shot and running the version 4 version 5 version 5.1 and version 5.1 raw all at the same time off of the same prompt The Prompt was cinematic still filmed by Stanley Kubrick Jack Nicholson as Napoleon Battle of Waterloo 35 millimeter Kodachrome so in case you didn't know that actually was supposed to be a real movie it's actually known as the greatest movie never made on a lot of lists so um I think it would have been pretty cool definitely better than Barry Linden which I do not like anyhow in order to add permutations all we do as at the end of our prompt we would put in curly bracket curly bracket dash dash V4 comma dash dash V5 comma Dash V 5.1 comma and then this is actually the tricky one is dash dash V 5.1 space Dash Dash style raw and then curly bracket close curly bracket close and once you hit it the prompt will then run once with each different version it was interesting to see how each version handled that prompt I think that it really showcased each of their strengths I will say that all of them missed the Jack Nicholson part of the prompt but when we've discussed this before I think that Napoleon combined with Battle of Waterloo comes in with such strong weight that it's gonna overpower Jack Nicholson I mean there are tricks with waiting Jack Nicholson and maybe multi-prompting but that's not what this example is about so I just decided to let it go circling over to illustrative Styles I did over the weekend Watch Dungeons and Dragons um so you know coming off of Kubrick I don't want anybody to think I'm super highfalutin or anything I watched Dungeons and Dragons it was awesome it's a really fun movie I do recommend it uh anyhow with d on my brain I just decided to try out digital painting Dungeons and Dragons ar169 with version 5.1 and these were the results which indeed has a dragon in a dungeon with a party of adventurers fighting it um very nice very colorful and very much in line with something that you would see in a Dungeons and Dragons book so I took that same seed and then ran it in version five and this was the output version 5 did switch some things around uh I do find that overall I think that first image and the third image are definitely the closest in terms of layout to the initial seed but overall you can definitely tell that there is a muting of colors in version five so once again taking that same prompt and seed number and running it in version 5.1 raw got us this which is closer to the initial layout of the seed but I feel really lacks a lot of details that were present in the initial V 5.1 output so overall I think that if you're looking for illustrative or Artistic Styles version 5.1 is definitely the way to go if you're looking for a photographic or cinematic type images I think that 5.1 raw is really what you should be looking at so I'm going to keep experimenting around with version 5.1 and version 5.1 raw please let me know what you've discovered about it as well in the comments below and from what I understand it's not going to be long before we're all back here talking about version six so that's exciting as well well I thank you for watching my name is Tim

Channel: Theoretically Media

Views: 65,399

Keywords: Midjourney, midjourney ai tutorial, midjourney ai, midjourney v5.1, midjourney v5.1 raw, midjourney v5.0, Midjourney tutorial, midjourney prompts, Midjourney lesson, midjourney update

Id: UacfIZavXYs

Length: 14min 41sec (881 seconds)

Published: Tue May 09 2023