Get The BEST of Both Worlds in DALL-E 3 & Midjourney

so in last week's video I talked about the announcement of do 3 and how it had not yet been released at least that's what I thought but it turns out I was wrong even though it's not available in chat TPT yet or it wasn't at the time of recording of this video Microsoft had already snuck it into the Bing image Creator so naturally I was really curious and I wanted to give it a try and I have to admit it's pretty godamn impressive as I already mentioned last week we kind of already knew that the image qual quality of Dolly 3 was considerably better than Dolly 2 but now that I've given it a try with my very own prompts I have to say it's much better than I expected but where Dolly 3 really excels is in its prompt coherence and I talked about this last week but boy oh boy the Improvement is absolutely huge I'm not going to say it's mind-blowing because honestly that word gets tossed around way too much and in six months time we're probably going to say this is not mind-blowing at all but I really got to admit this is on a whole another level now M's Aesthetics and image quality are still just a little bit higher than what Dolly 3 can produce nowadays but when it comes to prompt coherence Dolly 3 completely blows moury out of the water and let's face it if you're using moury professionally then we all know that this is one of M jour's biggest weaknesses now it's not really surprising that do has a much better understanding of prompts simply because open AI obviously as GPD 4 but M Journey has a lot of catching up to do now and I can really see a lot of people either cancelling their mid Journey subscriptions completely or at least downgrading it to a smaller plan the good news is that mid Journey finally has some proper competition and that can only be a good thing for us users because now we can actually leverage the strengths of each one of these tools in order to achieve the best outcome for our project so today I'm going to show you a workflow that leverages Dolly 3's Superior prompt coherence as well as mid Journey's just slightly Superior image quality all right so first of all I started off by creating an entirely new prompt in the Bing image Creator now normally if I were in mid journey I would be using a prompting style that is a blend between natural language and the use of some keywords and I even have a specific prompting framework that I use that I also go into in my mid Journey course but since it's reasonable to assume that Dolly 3 uses GPT 4 under the hood I simply used natural language like proper full sentences uh for writing this prompt so let me just quickly read out this prompt cinematic film still of a muscular Viking with long braided blonde hair wearing a horned helmet he's standing on top of a ledge overseeing a vast F he wields a sword with a massive blade that is stained by B battle now the most important words to keep in mind here are muscular long braided blonde hair horned helmet ledge vast shord wields sword massive blade and stain by battle now you might be wondering why I use stained by battle rather than Blood Stained or anything like that and that's because I want to use the exact same prompt within mid Journey as well and unfortunately if I do keep the word blood in the prompt in mid Journey then even if the does go through then it will be epimeral which means that it will simply disappear the next time I refresh the channel and I think we can all agree that this is one of the most annoying Parts about mour because you know you feel like you're being treated like you're in a kindergarten but even though we're all adults but you know that's just how it is anyway let's take a closer look at the images that Dolly 3 gave us I'm going to push myself up here just to make some space and I'll open up the very first one and you can clearly see that it has covered pretty much all of the aspects of the prompt and this is the part that is really really impressive you can see that you've got the blonde hair it's maybe not entirely braided in this case but you've got the horned helmet you've got the long sword that's stained you you've got a muscular guy he's standing on a Ledge that's you know right in front of the F so this is really really coherent and if I look at the other images it's it's also pretty much the same thing it's incredibly coherent and even the image quality I have to admit is really really impressive and if you don't see the images from M journey and from do side by side it's kind of hard to tell the difference but and you would you know I can see why people would say that this is really really detailed and on the same level but the Aesthetics are still slightly different that being said these images are absolutely stunning I have to admit and they are really really useful for a a lot of use cases I have to admit that I'm still pretty shocked just how close these are to the prompts because this is so far beyond anything you could do with moury in terms of prompting um and in terms of control but yeah just this is just what it looks like so now let's have a look at what this would look like if we do it in M journey and I'm going to quickly enter the very same prompt and just to be fair this is not how you would usually prompt a m Journey nowadays in order to get the kind of results that you want but since I want to keep things as simple as possible and as comparable as possible I'm just going to use the exact same prompt now I've done these exercises before so don't be surprised that I'm using a seed here I really just want to recover the exact same results as before to make sure that I can really illustrate to you what this workflow looks like anyway so let me generate these real quick all right and here's what they look like with M jour's default stylization you can clearly see that aesthetically and Visually these are far more dramatic right this is the main difference compared to what I've seen from Dolly 3 so far however as visually pleasing as they are they do miss the mark in certain regards so for example the hair is fine it's not really braided but I mean I'll let it pass but where's the helmet there's no helmet let alone a horned helmet I also am having trouble finding a sword it's more like I'm seeing a dagger and some other stuff the part that I do prefer about these images is how uh in certain images the Viking is overseeing the Fjord especially in the bottom left one so that's one part that I've noticed um The Journey seems to be doing better than Dolly 3 but if I'm really honest and I look at what I prompted Dolly 3 picked up on so much more of the elements and moury kind of missed well it didn't miss the point but it kind of missed a lot of important elements so if I wanted to get something really specific my journey would be tricky now before I move on let me quickly show you what this would look like if I used the raw uh style raw parameter and I'm just going to quickly generate these all right let's have a quick look at these and what you can notice is that there's a little bit less of that default stylization that you usually get with mid Journey if you're not using style raw um and again I like the overall composition of the top left and the bottom right one a lot CU it's pretty close to what I was actually thinking of um but again this this this Viking does not have a helmet it's not a horned helmet and you know I can't even find the sword in most of these so this is a bit disappointing to be honest it's not unexpected because I just know what Ma Journey does and the words are really really far back in the prompt and not towards the front so it's not surprising that they would not necessarily show up but still the fact that Dolly 3 can do this is really really really impressive now if we have another look at the dolly 3 images for most people this would be more than sufficient right this this is really really useful for most use cases but in some cases and especially in the creative industry you may want to use something that's closer to what M Journey did here simply because it's aesthetically so much more pleasing it's much closer to the type of color grading and the style that you would see in movies or TV shows not to mention the fact that Dolly 3 still only produces Square images so that's not particularly useful if you're working on something where you need a widescreen format so what if we wanted to Leverage The Prompt coherence of Dolly 3 and the visual style of mid Journey well we could start off by using the image that we created in dol 3 and use that as an image reference in a new prompt in m journei so let's do that so I'm going to quickly go back to Dolly 3 or the Bing image Creator and the the image that I'm going to pick is this one now I need to emphasize the fact that this image of the four images that were created is actually the one that is least coherent with the prompt but I'm going to take this one anyway just because I really really like this composition but just so you know the fact that the image is not 100% in line with the prompt is probably going to make things even more tricky for Mid Journey as well but let's go quickly back to my channel here and I'm going to enter my prompt again this is a prompt that I've prepared beforehand with a seed and what I'm doing is I'm using the image reference and the entire prompt that we used before except that we're adding an image weight parameter of two plus the seed let's quickly generate these Oh and before we have a look at these by the way you don't need to use the image weight of two you could use whatever you want 0.5 one or 1.5 it doesn't matter the reason why I'm using two is because I really want to shove that image reference down mid Journey's throat in order to get as close as possible to the original image reference reference but again you can do whatever you want so let's have a quick look at these images and what we can see is that they are still not entirely what we prompted but there's a couple things that are much better so first of all we can see that the hair is already slightly better there are some braids over here but most of all the Viking finally has a helmet and it's a horned helmet and it looks pretty cool actually and the Viking also had has at least in one of the images seems to have a sword maybe even in two of them it's not entirely sure what kind of Sword it is but it's pretty close and I do like the position that the Viking has in these images so he's really overseeing the F rather just than just standing on the side but I will admit that obviously this is not exactly the image that we used in the image reference but that's not necessarily the point here what we're trying to do is we're trying to get something that we couldn't have prompted directly in M Journey and get it from Dolly 3 reinjected into mouri hoping to get closer because afterwards we want to use the visual style like the strengths of mouri to create something else so even though these four Images are already pretty decent and they've Incorporated a lot of the elements that weren't there before they still lack texture and depth luckily there's a really simple way to fix this so first of all we need to upscale the image that we want to keep and I'm going to choose the upper right one sorry top right one I'll hit the U2 button and I'll wait for that to upscale and once that upscales I hit the very subtle button that will open up the remix prompt and then I simply remove the image reference from the remix prompt as well as everything here at the back and then I simply H submit and what this usually does it simply adds more detail more texture more depth as well as better color grading to the image hey sorry for interrupting you right in the middle of the video but I just wanted to share something with you as many of you know I have an online video course called masters of M Journey which teaches you all of the foundational skills that you need in order to get the most out of M journey and I know that some of you are still on the fence trying to figure out whether this is right for you and I get it you look at all of the free stuff that I share right here on my YouTube channel and you ask yourself how much more can there really be well many of the things that I do in these videos including the method that I just used are explained in far more detail in the course I show many different examples and I teach these to a degree that make sure that you really understand how they work and you can apply them to pretty much any project not just one particular use case so if you want to learn how to control mid journey and prompt with intent then visit Masters ofm journey.com I'll see you on the flip side it also just gives you a couple more options to choose from without actually breaking the image itself and now from these images you would simply pick the one that you like the most you would upscale it and I'm going to pick the bottom sorry the top left one and then once that's upscaled then I can do additional operations that I would not be able to do within Dolly 3 but which I can do in mid Journey so for example I'll hit the custom zoom button in order to add more context to the image and before I hit submit I'm still going to change the zoom factor to 1.5 and I will also change the aspect ratio 26 to9 that I hit submit I'll wait the for these to generate and then we end up with this set of four images of which you can simply pick the best one which I personally think is this one which you can see on screen right now now while this particular example isn't necessarily perfect and your results May Vary depending on the topic that you're working on I think it is a useful workflow for combining the strengths of both tools so now let's move on to another example so let me quickly go back to the Bing image Creator and for this example that I'm about to show you again I used 100% natural language as if I was writing Pros for a text for a I don't know story it doesn't matter I just use regular language cinematic film still of brunette with braided bun hair in her 30s wearing a black wool trench coat over an ochre turtleneck and Je means she's walking through the streets of M on a sunny autumn day enjoying the Vivid street life and the specific words to look out for here are brunette braided buir black wool trench coat ochre turtleneck jeans streets of MRA Sunny autumn day and street life and now let's check out what Dolly 3 produces based on this prompt and if we look at these images just going to open up the first one you could see that these are in incredibly coherent so first of all we've got the brunette we've got the braided bun here which is pretty special that's shown up in this image um I can't really tell whether she's in her 30s or not but let's just assume this looks like her 30s then the black wool trench coat which is quite obvious the ochre turtleneck and the jeans are not that visible but there's a hint of jeans and she's walking through the streets that I can attest to look very much like and it's a sunny autumn day and there's street life now let's have a look at the other images just to show you just how coherent this is it's it's really quite shocking okay I'm I'm I'm very pleasantly surprised to see not just the image quality which is very very high but also just how detailed and how specific it is to the prompt and I'll show you the third one and the there we can see the jeans by the way and the fourth one now the only thing that bothers me about these images is that all of the faces look a little bit fake now I know some people might say no it's not fake she's just pretty but no to be honest these faces they look a little bit like I don't know like she had some work done like I don't know what it is nowadays but it's it's just not the kind of natural this one's probably the most natural one but the other ones look a little bit like she's had a nose job or something like I don't know whatever and every time I prompt this it's pretty much similar so this kind of begs the question what mid Journey would give us using the exact same prompt so let's give that a try I'm going to go over to Mid Journey now and I'm going to Simply enter the prompt just like I did before I'm going to use the exact same prompt as for Dolly 3 I'm I'm going to use style raw here and also a seed simply because I prepared this and let's generate these all right so let's have a quick look now remember I said that I use style raw here but I can tell you that even with the Basse sty ation of moury they look pretty similar they're just as natural and that's one of the things that you can clearly see in these images the they they're just a lot more natural than the images produced by Dolly 3 now you may like that or you may not like that but it is a very specific visual aesthetic now the other thing is though that you'll see is that while mouri has managed to pick up on a lot of the elements it's also missed out on a lot of the elements so she is mostly wearing a black wool trench coat she's also wearing a turtleneck that is either ochre or in some of these images is going more towards a brown tone but I'll let that pass um that being said the hairstyle is definitely not a braided bun hairstyle um she is a brunette which is fine she's probably in her 30s she looks like that and the surrounding environment is very much M you can even kind of of see the Eiffel Tower there in the back I think in in in the top right image um even though it's a bit pointy and there is street life now you can also debate on whether it's sunny enough I think this is actually more realistic for a sunny autumn day while the other one looked like it was a bit too sunny but again these are these are semantics these are details um I'm not going to argue about that so now let's check out whether we can use one of the dolly 3 images and in this case I'm going to use not this one but this one which I like the best in order to create an image in mid Journey which has much more Fidelity to The Prompt so I'm going to go back to M journey and I'll re-enter a new prompt imagine and I'll add the image reference at the very front and again I'm using an image weight of two style raw and again I'm using a c just so I can recover whatever I worked on before I'll generate these real quick all right so these results were actually already quite interesting thing just by using the image reference we've managed to inject a lot of the elements of do the dolly 3 image so first of all the woman definitely has the right hairstyle right hair color the the whole outfit is absolutely perfect except for maybe the genes I can't really tell whether they're genes if they're genes they must be pretty dark and the the overall surrounding environment is much closer to what M Journey produced and the face is already much more natural than in the dolly 3 images yet it still does look a little bit more like she was prepared for a photo shoot or something like that so but I what I really find fascinating here is the turtleneck cuz the turtleneck has just really really nice texture this is really really great already so what I'll do next now is I'm going to pick one of these images again just like we did before with the Viking and in this case I'm going to take the top left one because I like that composition and I'm going to upscale it and once it's upscaled I'll hit the very subtle button again this will open up the remix prompt I will remove the image reference and I will remove everything else here at the back as well maybe except for style raw I'll leave that in and then I'll hit submit all right and what we get are these four Images which as you can see are considerably more natural still than the others of course there's been some dilution of the elements but still all of the key stuff is there and it doesn't really matter whether you like one or the other some people may prefer the early one whereas others will enjoy this one that's not really the point the point is to show you that you can do this and you can combine elements from Dolly 3 with mid journey and you can just combine those strengths and if you wanted to you could now take one of these images expand the context with custom Zoom or by panning or what you could also do is you could let's say we upscale u4 here and then let's say we wanted to have this woman this particular woman with the same facial consistency in slightly different camera angles and perspectives so what we would do then is I would first copy the link of the image and by the way if you're not familiar with this toolbar down here this is my prompt aot Chrome extension which you can get at prompt alot.com extension and it helps you a ton by you know just making it more efficient and easier to copy and paste different parts of the prompt I really recommend getting it it's free of charge for the bass version and yeah that's about it now I've copied the image link and now I can hit the very strong button and then I'm going to reinsert the image reference here at the front in the remix prompt simply because otherwise there's a risk of losing more of the key elements of the original image so for example the braided bun here that's an element that was injected by the image reference from before and not by The Prompt so if we leave this out then the fact that the text prompt doesn't really represent present that little detail will most likely dilute it away so that's the reason why I'm putting it in here that's that's really just the only reason I hit submit and as you can see we now have four additional images of the very same woman with the exact same facial features same outfit same hairstyle in more or less the same location but different shots anyway I hope you found these insights useful and maybe you can integrate these into your own workflows as well that's it for today remember to check the video description for a whole bunch of free stuff including my Chrome extension as well as my mid Journey course and keep on learning and take care [Music] cheers

Channel: Tokenized AI by Christian Heidorn

Views: 12,700

Keywords: midjourney, midjourney v5, midjourney version 5, midjourney tutorial, midjourney prompts, midjourney workflow, midjourney learn, midjourney 5.2, midjourney 5.3, dall-e, dalle 3, dall-e 3

Id: uCIhb4vLd2I

Length: 22min 55sec (1375 seconds)

Published: Sat Oct 07 2023