Discover Midjourney V5's Hidden Powers with GPT-4!

Another massive release just happened: Midjourney  V5 just came out. And I already found some tricks   to get better results. But why would you even want  to use this? And once you use it, what do you need   to look out for? We're going to be discussing  all that, plus how to use GPT-4 to get superior,   majority V5 results today. So let's get into it! So first up, I need to tell you that only paying   mid-journey members have access to this.   But if you're a paying subscriber starting   at $10 a month, it's quite easy. You just go  into Discord, say "settings," and now you can   change your model. This is the most comfortable  way to interact with it. Otherwise, you need to   include "--V Space 5" at the end of your prompts. But why would you want to use it? Well, simply

put, the biggest reason to use V5 against V4  right now is a desire for more realistic images.    Concretely, humans. They finally fix the hand  situation, so sometimes you still get an extra   finger, but it's not a catastrophe every single  time anymore. Also, the photorealistic qualities   of this thing just went through the roof. These  pictures look as if they were taken with a camera,   like they're 95% there, with great skin texture  and proper hands, and even details like teeth   now being properly represented because they train  this new model to do exactly that. You can start   using this thing to replace models right now. But the more realistic settings that are baked   into the system are not everything here.   Overall, you can view V5 as more of a pro   model than V4 was. You need to go more into detail  to get better results. So as opposed to GPT-4,   where it got smarter and it got a little easier  to prompt, this moved in the other direction. You   can get better results, but your prompts need to  be better. And here's one of the biggest changes:   no more prompting with keywords only. At least,  they don't recommend it. You should use natural   language, just like I'm speaking to you in  right now. Just like you would communicate   with a human. If you just take a step  back and think about this for a second,   you'll realize that natural language includes so  much detail and so much direction that obviously,   it's going to be able to craft better results  from it than language including details better.    All those little words in between actually do a  lot to communicate what we want to communicate.    So no more keywords only. But once I show you the  prompt that I crafted in GPT4, you'll still be   able to use keywords, because GPT4 is gonna do the  natural language processing for you, but more to   that soon. Another change that I personally really  enjoy is that it's upskilled already. So no more   waiting twice if you want to upskill something. So  if it just generates a cat with a hat right here,   right away, you should be able to see how  realistic these are by default without me   including anything like photorealistic or  aperture. All I asked for is a cat with a   hat and I got photos of a cat with a hat. So yeah,  I just realized, this photo-realistic capabilities   are baked in now. And, as I mentioned, the  upscaling is super fast. So, in real time, I'll   just click U1, meaning upscale one, and without  even finishing my sentence, the thing is done.    This looks way too real! If you were to  show this to a normal person, I think this   passes as a photo, as long as a person is not  consciously looking for AI-generated features.    With my photography background, I would say the  depth of field gives it away a little bit. These   whiskers should have been sharp because the  ones on the left and the eyes are sharp too,   and they're in between those two focal planes.   But that's just me nerding out. Now let's move

on to the interesting part, and that is how  to get better results by using GPT-4 for this.   So, what I did here is I prepared a little prompt  for you guys. What you want to do is you want to   go to GPT-4. As a result, the outputs for this  are slightly better than 3. 5 because we're kind   of pushing it here with the amount of detail  we request. You will be able to simply copy   paste this from the description of this video,  but essentially, it makes GPT-4 act as a stable   diffusion photography prompt generator. And then,  we specify the input and the outputs. So, we input   the visual description, and it outputs a detailed  paragraph that I can copy into my diffusion model.    Include a variety of photography-related  terminology, including the description   of the exact Zeiss lens. So, I specified a brand  here because that's my favorite look, and that's   also a lot of the equipment they use around here  because I just love that sharp look. It looks very   professional and very premium, very luxurious, and  this is what I want for my images too. But you can   switch this up to any brand. You could even say  iPhone lens to imitate photos a normal person   would take in their everyday life. Either way, the  prompt continues, and most importantly, a detailed   description of the volumetric lighting. So, if you  didn't know, volumetric lighting is a term that I   borrowed from my background in videography and  cinematography, and it's what some people refer   to as God Rays. Essentially, this diffused beam  of light in the background of the image. Again,   you can change this for a different type  of lighting, but I really enjoy this in my   images. As in combination with the Zeiss lens, it  makes it epic and luxurious, something I really   enjoy in my images. And then we end with, now,  right near clear and concise natural language   visual description of the following. The natural  language part here is very important because,   as I said, it doesn't want keywords anymore.   But the cool thing is, I could go ahead and   use keywords to describe this now, because I  have a prompt generator. So let's do this. And,   my idea here is, let's generate some prompts that  recreate items in my room. For example, there's   the ship inside of a bottle right on my table. So  let's try and recreate the epic version of this,   inside of Mid Journey V5. Okay, so all I'll do  here in the end is, a ship inside of a bottle,   and say 'Enter. ' That's simple. And what it does  is create a detailed description of my scene,   including a specific lens. And I'll squeeze  in this little tip: the lens here has become   increasingly important, as these outputs are  very realistic. Just like in photography, one   of the main things that influences that look is  your lens. So, there never has been a better time   for you to learn a little bit about photography.   And although I would love to sit for a few hours   and give you a basic lecture, because still,  I'm so passionate about all this camera stuff,   I'll make it simple for you. And I included two  resources in the description that will give you a   basic overview of lenses. Okay, so look over here.   Sigma is one of the main manufacturers of lenses,   and at this fantastic page, which shows  off different focal lengths and full clips,   are essentially how wide your shot is. That's  all you need to know. And whenever you want a   specific look, you can simply look at this  page and pick this millimeter number right   here and switch that out inside of your prompt. Now, if you want to get even more advanced,   you could pick specific lenses. And a good  way to do this is by going to the website of   B& H Photo Video. com. When you go to Photography  and Lenses, you can pick from all the different   brands. So just pick SLR lenses because there's  the most variety there. And then we'll do   something most people won't do: list from high  to low. We want the most expensive ones, right?   And right here, you already see that it actually  shows the Zeiss Auto's lenses which GPT4 picked   by default. I mean, how smart is that? And  the Otis lenses are very well known for   portrait photography. By many people, they  are considered the holy grail of portrait   photography because they're super expensive,  super sharp, and GPT4 already uses them.   But you could just look at this bundle, and here  in the description, you have all the lenses in the   bundle. So I could just copy this 100 millimeter.   And again, here I would just check, "Ah, 100   millimeters, really zoomed in? Yes, that's kind  of what I want with my ship inside of a bottle.

Another massive release just happened: Midjourney  V5 just came out. And I already found some tricks   to get better results. But why would you even want  to use this? And once you use it, what do you need   to look out for? We're going to be discussing  all that, plus how to use GPT-4 to get superior,   majority V5 results today. So let's get into it! So first up, I need to tell you that only paying   mid-journey members have access to this.   But if you're a paying subscriber starting   at $10 a month, it's quite easy. You just go  into Discord, say "settings," and now you can   change your model. This is the most comfortable  way to interact with it. Otherwise, you need to   include "--V Space 5" at the end of your prompts. But why would you want to use it? Well, simply   put, the biggest reason to use V5 against V4  right now is a desire for more realistic images.    Concretely, humans. They finally fix the hand  situation, so sometimes you still get an extra   finger, but it's not a catastrophe every single  time anymore. Also, the photorealistic qualities   of this thing just went through the roof. These  pictures look as if they were taken with a camera,   like they're 95% there, with great skin texture  and proper hands, and even details like teeth   now being properly represented because they train  this new model to do exactly that. You can start   using this thing to replace models right now. But the more realistic settings that are baked   into the system are not everything here.   Overall, you can view V5 as more of a pro   model than V4 was. You need to go more into detail  to get better results. So as opposed to GPT-4,   where it got smarter and it got a little easier  to prompt, this moved in the other direction. You   can get better results, but your prompts need to  be better. And here's one of the biggest changes:   no more prompting with keywords only. At least,  they don't recommend it. You should use natural   language, just like I'm speaking to you in  right now. Just like you would communicate   with a human. If you just take a step  back and think about this for a second,   you'll realize that natural language includes so  much detail and so much direction that obviously,   it's going to be able to craft better results  from it than language including details better.    All those little words in between actually do a  lot to communicate what we want to communicate.    So no more keywords only. But once I show you the  prompt that I crafted in GPT4, you'll still be   able to use keywords, because GPT4 is gonna do the  natural language processing for you, but more to   that soon. Another change that I personally really  enjoy is that it's upskilled already. So no more   waiting twice if you want to upskill something. So  if it just generates a cat with a hat right here,   right away, you should be able to see how  realistic these are by default without me   including anything like photorealistic or  aperture. All I asked for is a cat with a   hat and I got photos of a cat with a hat. So yeah,  I just realized, this photo-realistic capabilities   are baked in now. And, as I mentioned, the  upscaling is super fast. So, in real time, I'll   just click U1, meaning upscale one, and without  even finishing my sentence, the thing is done.    This looks way too real! If you were to  show this to a normal person, I think this   passes as a photo, as long as a person is not  consciously looking for AI-generated features.    With my photography background, I would say the  depth of field gives it away a little bit. These   whiskers should have been sharp because the  ones on the left and the eyes are sharp too,   and they're in between those two focal planes.   But that's just me nerding out. Now let's move

on to the interesting part, and that is how  to get better results by using GPT-4 for this.   So, what I did here is I prepared a little prompt  for you guys. What you want to do is you want to   go to GPT-4. As a result, the outputs for this  are slightly better than 3. 5 because we're kind   of pushing it here with the amount of detail  we request. You will be able to simply copy   paste this from the description of this video,  but essentially, it makes GPT-4 act as a stable   diffusion photography prompt generator. And then,  we specify the input and the outputs. So, we input   the visual description, and it outputs a detailed  paragraph that I can copy into my diffusion model.    Include a variety of photography-related  terminology, including the description   of the exact Zeiss lens. So, I specified a brand  here because that's my favorite look, and that's   also a lot of the equipment they use around here  because I just love that sharp look. It looks very   professional and very premium, very luxurious, and  this is what I want for my images too. But you can   switch this up to any brand. You could even say  iPhone lens to imitate photos a normal person   would take in their everyday life. Either way, the  prompt continues, and most importantly, a detailed   description of the volumetric lighting. So, if you  didn't know, volumetric lighting is a term that I   borrowed from my background in videography and  cinematography, and it's what some people refer   to as God Rays. Essentially, this diffused beam  of light in the background of the image. Again,   you can change this for a different type  of lighting, but I really enjoy this in my   images. As in combination with the Zeiss lens, it  makes it epic and luxurious, something I really   enjoy in my images. And then we end with, now,  right near clear and concise natural language   visual description of the following. The natural  language part here is very important because,   as I said, it doesn't want keywords anymore.   But the cool thing is, I could go ahead and   use keywords to describe this now, because I  have a prompt generator. So let's do this. And,   my idea here is, let's generate some prompts that  recreate items in my room. For example, there's   the ship inside of a bottle right on my table. So  let's try and recreate the epic version of this,   inside of Mid Journey V5. Okay, so all I'll do  here in the end is, a ship inside of a bottle,   and say 'Enter. ' That's simple. And what it does  is create a detailed description of my scene,   including a specific lens. And I'll squeeze  in this little tip: the lens here has become   increasingly important, as these outputs are  very realistic. Just like in photography, one   of the main things that influences that look is  your lens. So, there never has been a better time   for you to learn a little bit about photography.   And although I would love to sit for a few hours   and give you a basic lecture, because still,  I'm so passionate about all this camera stuff,   I'll make it simple for you. And I included two  resources in the description that will give you a   basic overview of lenses. Okay, so look over here.   Sigma is one of the main manufacturers of lenses,   and at this fantastic page, which shows  off different focal lengths and full clips,   are essentially how wide your shot is. That's  all you need to know. And whenever you want a   specific look, you can simply look at this  page and pick this millimeter number right   here and switch that out inside of your prompt. Now, if you want to get even more advanced,   you could pick specific lenses. And a good  way to do this is by going to the website of   B& H Photo Video. com. When you go to Photography  and Lenses, you can pick from all the different   brands. So just pick SLR lenses because there's  the most variety there. And then we'll do   something most people won't do: list from high  to low. We want the most expensive ones, right?   And right here, you already see that it actually  shows the Zeiss Auto's lenses which GPT4 picked   by default. I mean, how smart is that? And  the Otis lenses are very well known for   portrait photography. By many people, they  are considered the holy grail of portrait   photography because they're super expensive,  super sharp, and GPT4 already uses them.   But you could just look at this bundle, and here  in the description, you have all the lenses in the   bundle. So I could just copy this 100 millimeter.   And again, here I would just check, "Ah, 100   millimeters, really zoomed in? Yes, that's kind  of what I want with my ship inside of a bottle.   And once I copy my prompt into my journey  again, I'll just say, I'll just say:   /imagine as per usual. I could just go ahead and  copy the name of this 100 millimeter lens, and   that's the only thing I'm going to change about  my prompt. Okay, and now I'm gonna hit enter.    So now it should be less wide  and more focused on the details,   which is exactly what I want with something  as detailed as a ship inside of a bottle.   Now, all we have to do is wait for a minute and  look at the results. Will you look at that? I   mean, that's pretty impressive. Look at that. That  looks so realistic. This could easily be a real   photo. Now, given I could have included a lot of  specific details to make it look just like this,   but honestly, I just wanted to see what  my journey plus GPT-4 would come up with.   Okay, and to round this out, I just want to do  one more because I know in V4, this prompt always   generated an image of something very cartoonish  looking. I'll just say: "AI YouTuber at his desk. "  And funnily enough, it actually uses a lens  that I typically have on top of this camera,   the Zeiss, but I usually have the 40 millimeter.   Oh, it really went hard on this description, so   let's see what we get here. And I'll just switch  this to the 40 millimeter And hit enter! Oh wow,   look at this. Okay, given there are many  details which are messed up, like what is   like my journey, develop some camera prototype  that I've never seen before, but look at this   guy sitting at his desk, and the hands are  not good, but they're also not terrible. They   used to be just terrible, right? Okay, so that's  it. If you create some amazing images yourself,   I want you to join our Discord server and share  them with the community. Me and many others in   there are posting their creations and discussing  what is going on in the AI space on a daily basis.    And if you want to get even better at Midjourney,  you should check out this video because it will   teach you the ultimate hack on how to get the most  out of it with just one keyword. See you there!