https://www.krea.city/articles/lintelligence-artificielle-generation-contenu-texte-image

I use Midjourney to create characters, icons, stickers, a bit of everything.

And here I'm going to show you 3 aspects of Midjourney that you may not know yet because it was released recently.

We're going to talk about character generation, so trying to have a consistent character but in several different environments.

We will talk about zoom, pan and different small options that have arrived in Midjourney 5.2 and then we will see how we can download a whole series of files, a whole series of images in its Midjourney image gallery to download our character that we just created in 10 copies, in 10 different versions to download all this in 1 go.

Hello, my name is Vince and welcome to CREA-CT.

We will do these 3 things, we will create a character that will be a little bit the same everywhere in several different images.

So we're going to try to create a series of characters that are a little bit the same, but in different environments, different poses, etc.

And then we're going to think about this whole thing of zoom and pan, see what it can do.

And we will end with a little tip for your gallery here.

Because at 1 point I wanted to download everything, I didn't know how to do it exactly.

To select everything and download all the images at once.

So I found a little tip and I will show you that in third place First thing, we will start by creating a character that we will try to create as consistent that is to say the same character but in many different environments you see that Here I have different characters that I placed in different environments at 1 time or another.

It was for a video that will be released soon, and it's called Julie makes no code, so I needed Julie, so I created Julie.

You see, the same lady, but placed in several environments, in several ways.

I also did tests here with the same character, but with different views and different angles.

And so we're going to use these little arguments, different angles and character sheet.

Here I will start with a prompt which will be imagine.

We will take something very simple boy dark air different angles and character sheet ok we will start with something quite simple just with that to see a little bit what it gives and we will already realize what it can offer us as a possibility.

You can see that we have the same character taken several times.

But I'm going to take this 1 and I'm going to make it bigger.

So U2 to have this 1 bigger and full size.

Meanwhile, I remove this 1, I don't need it anymore, consequence the little cross and it deletes.

Now, starting from there, I would like more characters.

So obviously I could do variations, a remaster, but I will first start by using 1 of the new features of Midjourney 5.2, it's zoom out.

And so zoom out, in fact, I will enlarge the image and it will create more characters around those that are already created.

Ok, so I'm back now with this sheet, this series of images, these different possibilities that are offered.

I'm going to let you now fiddle with it and see what you can get but in any case you have a small base of reflection for that So let's try now with a photorealist character so here I simply put photo, reel, a man, dark air, blue eyes.

What we're going to achieve with this kind of image is a little different from what I did at the beginning.

Now that we have this character, we can start doing several things.

The first thing we can do is to zoom in twice, and then a small pan to the left, a small pan to the right to see what it gives.

At the level of the pan out, we have things that are quite interesting in my opinion, and that are pretty good.

There is none of the 4 images that is really completely bad.

But when we start to do the rest, Midjourney tries to complete the image as best he can.

I still like this 1, it seems pretty clean to me.

To add text here, for example, in the negative space, it seems quite interesting to me.

And if we zoom out of this 1, it should be pretty good.

Besides, that's what we're going to good.

We have just a zoom out, so from there, maybe zoom back again if possible.

So we're going to start again from this 1 which is interesting, the third 1 and we will maybe go back a little bit so this 1 seems a lot more complicated there is not much to recover in my opinion except maybe the fourth 1 that we can try to see in bigger Those that are set to pan up, well, we just went up a little bit, but somewhere it also returned to a square mode.

I'm not sure there's anything to pick up.

Here, here, pan down, yeah, ok, good, not too bad, but I'm not sure it has a striking interest in this image.

The variations, on the other hand, indeed, between variation and remaster, there is always this tension, this interest to ask yourself what is the point of having 1 or the other.

1 that is rather automatic and the other for which we can add additional parameters in any case.

So that makes the difference between the 2.

On the other hand, in the remastered version, it will look for different propositions the eye color is not there anymore, it's starting to become a bit extreme maybe this 1 is the most valid, here at the top right but remastered is maybe a bit extreme as a solution Pan down, once again we are below So, on the other hand, we are pretty good I think Because now here we have a nice image With a negative space here The goal was to create characters and variations of characters a little different and so I will try to recover here an image as complete as possible a rather interesting image this 1 still seems to me not bad and here I will recover the address of the image copy image address hop here and I will come to reuse it here and I will do imagine I will paste the address of the image and then what I will

going to do is create a little game of variables.

So I'm going to say, angry, joyful, fearless, for nothing.

And then I'm going to close this, reopen this, and I'm going to say, maybe, very bright environment, at the beach, full sun.

Another environment, at the office during meeting.

And we're going to create like that, and I have to close with a close and not a parenthesis, otherwise it won't work.

So, by taking the image I found the link that was there, I added 3 options and 3 options, which means that it will give me 9 image possibilities.

Yes, I want to create all these variations, all these combinations.

So what did that give us as a result of this story?

Well, if we look a little bit, here it is, it may not be famous.

I like this image, if in reality I wonder if it's even at work or rather a police interrogation, but let's say I like this 1 pretty well as an image, I think.

Here we have things that are much stranger, sometimes.

I wonder how the algorithm does its job and how the thing creates that.

I'm not very sure, but let's say I like this image anyway.

It makes me laugh, the image here on the top right.

So here, on the other hand, it's irrecoverable, I think, because at the same time, our guy, he lost 15 years, but he kept his beard.

That is to say that here, this childhood, it seems a little strange to me.

It's not going to be possible, I want to get something out of it.

Where it's going to be more interesting, it's here.

I like it here, I think there are some good images in there.

I would say that the 1 that looks the most like our original subject is perhaps this 1.

The beard was a little longer but we still keep a certain character.

Here maybe not, here it's not quite the same head I think.

So let's say we keep the 1 below here I think.

So there, on the other hand, he's gone into a total delirium.

And I like the image here below, somewhere.

We're not in photorealism at all, but why not?

I don't know, I don't see what we can keep.

It's a desk at the same time and a plantation at the same time.

So, I like this image, I think there is a certain light, a certain nice thing.

So here we are really very very beautiful, so this 1 here at the top right, it is just beautiful.

Because, I don't know what to say, but...

I don't see what we're going to do with it.

Apart from laughing, it could be nice if you have someone who looks like this at work.

You take the image, you recover it, you punish it on the board, it's not very nice, but...

But somehow it depends on the environment.

This 1 looks like the short person in Game of Thrones.

You have all the people in the office watching you, and you're still stuck in yourself.

I think this image tells a story, there's a way to make it something interesting, in my opinion.

So I put it in variations to see if we can get better.

This 1 I think is pretty good, it's really nice and it's the same character as earlier, so we have something that is good.

This 1 is already a little more complicated but we are maybe a little more in Fight Club mode.

This 1 in my opinion corresponds quite well.

This 1 I told you what I thought about this 1, I think it's nice, but in another genre eventually.

This 1 is not bad, but it's not quite the same character, it's a bit too smooth here to be exactly the same character.

I find this 1 really good in a certain context to tell a certain story, I find it quite nice.

So here we are, because it was the goal of the video anyway, it's to show you how to create coherent characters, characters that always remain a bit the same.

If we take the character we had at the beginning and with which we started earlier, this 1, there is indeed a form of resemblance, it has moved away a bit, but it does the trick, it is probably enough compared to what we would like to tell as a story, or if it's images for social networks or something like that.

Obviously, we're not going to go with that in production mode, and we're not going to cover a magazine with that kind of image, but in the illustration of a sentence or in the way of saying things on social networks.

For example, I find that this image with a little text here in the negative space here, I think there is still something to do, I think it's quite interesting.

Same thing for this 1, somewhere there is probably a way around this character to find something to tell.

And then there are aberrations or things that make no sense and are completely off the mark.

It's free to generate more images in a way.

Me clearly here, the video I'm shooting now, it's been shooting for 48 minutes, to generate images, it does take a while, and if we do nothing else during that time, we still lose a lot of time, a lot of hours, a lot of minutes in these creations.

What I advise, if there is a way, is to have another computer or a tablet on the side that is made by the generation.

And during this time, we continue to work, we continue to do other things, because otherwise we really lose a lot of time.

But when I see the images that we get in total, I tell myself that it's not so bad.

Here, in any case, in terms of generating a character that looks like the original character.

By playing a little more with the contrasts or with the texture, whether it's in Canva, Photoshop or the software you use for image processing, I think we can get something that is still quite interesting.

The last thing I wanted to show you, beyond the character generation, the pan, the zoom, the x2, etc.

Is how we get by here with our gallery when we want to select several images.

So here I'm going to do a little reload so that it recovers the last images we just created.

And so I say to myself, OK, I'm going to get all the images of my character here to make a zip and to be able to use it.

Well, now here we have this option to select and choose the images we want to keep.

So if I do that, be careful to click well outside the area that just added there.

And then at 1 point or another, it will simply create a zip with my images and we will do the downloader here so I will do this prepare a zip we will say here that we do not need to make a huge batch size but I have already downloaded 1050 images like that in 1 go and he created me several packets several zips of 250 images I tried with 1000 it didn't work so here it is and so I'm going to download all this he's going to put the images in a zip to retrieve the zip We have our images that have been downloaded and now we can use them in posts and things like that.

So here's the result, even a little enlarged, I still find it quite nice, quite clean.

I think there's a way to do something quite interesting with that on social media in particular.

So on that, I think the main thing is go ahead, create, show me what you do.

I want to see a little bit of your different creations.

I am very tempted by the story of Stable Diffusion and the story of creating his own model to generate images that are pretty good in the vein of what I would like to do with CreaCity.

I know that Thomas Boyle does some pretty incredible things with this story and I would be really super tempted to try to create the world of CreaCity, but thanks to Stable Diffusion and a model that I will train myself to generate images that I want to see, and not just the images that Midjoney thinks are interesting.

And then, I think there will be a lot of things to do.

Here I will soon go on vacation and, in my opinion, generate images.

I will generate some during the holidays, because it's the kind of thing you throw on the tablet you generate some images, you keep playing cards and then you come back and you look at it from time to time it's even funny to show the children and the parents I find, we show them a little bit what the engine has created, among other things the image of the guy with his grunt there, it was still quite disturbing.

So here it is, once again my name is Vince and thank you for watching this video.

Don't forget to like, comment, share, subscribe and ring the bell.

Be nice in the comments because trolls like to know when they dance, sing or give hugs.


Speaker 0: As you probably know, I love Midjourney. I use Midjourney to create characters, icons, stickers, a bit of everything. And here I'm going to show you 3 aspects of Midjourney that you may not know yet because it was released recently. We're going to talk about character generation, so trying to have a consistent character but in several different environments. We will talk about zoom, pan and different small options that have arrived in Midjourney 5.2 and then we will see how we can download a whole series of files, a whole series of images in its Midjourney image gallery to download our character that we just created in 10 copies, in 10 different versions to download all this in 1 go.

Hello, my name is Vince and welcome to CREA-CT. We will do these 3 things, we will create a character that will be a little bit the same everywhere in several different images. So we're going to try to create a series of characters that are a little bit the same, but in different environments, different poses, etc. And then we're going to think about this whole thing of zoom and pan, see what it can do. And we will end with a little tip for your gallery here.

Because at 1 point I wanted to download everything, I didn't know how to do it exactly. To select everything and download all the images at once. So I found a little tip and I will show you that in third place First thing, we will start by creating a character that we will try to create as consistent that is to say the same character but in many different environments you see that Here I have different characters that I placed in different environments at 1 time or another. It was for a video that will be released soon, and it's called Julie makes no code, so I needed Julie, so I created Julie. You see, the same lady, but placed in several environments, in several ways.

I also did tests here with the same character, but with different views and different angles. And so we're going to use these little arguments, different angles and character sheet. We will start and we will launch. Here I will start with a prompt which will be imagine. We will take something very simple boy dark air different angles and character sheet ok we will start with something quite simple just with that to see a little bit what it gives and we will already realize what it can offer us as a possibility.

Here's what it looks like. So I'm going to take this series here. You can see that we have the same character taken several times. This 1 is not bad either in a way. But I'm going to take this 1 and I'm going to make it bigger.

So U2 to have this 1 bigger and full size. Meanwhile, I remove this 1, I don't need it anymore, consequence the little cross and it deletes. Now, starting from there, I would like more characters. So obviously I could do variations, a remaster, but I will first start by using 1 of the new features of Midjourney 5.2, it's zoom out. And so zoom out, in fact, I will enlarge the image and it will create more characters around those that are already created.

Ok, so I'm back now with this sheet, this series of images, these different possibilities that are offered. So everything is not perfect, of course. I'm going to let you now fiddle with it and see what you can get but in any case you have a small base of reflection for that So let's try now with a photorealist character so here I simply put photo, reel, a man, dark air, blue eyes. You have to choose between these 2. I'm going to take the 1 below.

What we're going to achieve with this kind of image is a little different from what I did at the beginning. Now that we have this character, we can start doing several things. The first thing we can do is to zoom in twice, and then a small pan to the left, a small pan to the right to see what it gives. At the top. At the bottom.

Mix square, a variation and a remaster. And see what it gives. So what do we have as a result? At the level of the pan out, we have things that are quite interesting in my opinion, and that are pretty good. There is none of the 4 images that is really completely bad.

I think they are all 4 quite valid.

Speaker 1: As a result, the pan out is quite good. None of the 4 images are bad. I think

Speaker 0: all 4 are quite good. But when we start to do the rest, Midjourney tries to complete the image as best he can. I still like this 1, it seems pretty clean to me. To add text here, for example, in the negative space, it seems quite interesting to me. And if we zoom out of this 1, it should be pretty good.

Besides, that's what we're going to good. I'm going to launch this 1 here. We have just a zoom out, so from there, maybe zoom back again if possible. So we're going to start again from this 1 which is interesting, the third 1 and we will maybe go back a little bit so this 1 seems a lot more complicated there is not much to recover in my opinion except maybe the fourth 1 that we can try to see in bigger Those that are set to pan up, well, we just went up a little bit, but somewhere it also returned to a square mode. I'm not sure there's anything to pick up.

Come on, we're eliminating. Here, here, pan down, yeah, ok, good, not too bad, but I'm not sure it has a striking interest in this image. So, the variations. The variations, on the other hand, indeed, between variation and remaster, there is always this tension, this interest to ask yourself what is the point of having 1 or the other. 1 that is rather automatic and the other for which we can add additional parameters in any case.

So that makes the difference between the 2. I think he did a good job there. On the other hand, in the remastered version, it will look for different propositions the eye color is not there anymore, it's starting to become a bit extreme maybe this 1 is the most valid, here at the top right but remastered is maybe a bit extreme as a solution Pan down, once again we are below So, on the other hand, we are pretty good I think Because now here we have a nice image With a negative space here The goal was to create characters and variations of characters a little different and so I will try to recover here an image as complete as possible a rather interesting image this 1 still seems to me not bad and here I will recover the address of the image copy image address hop here and I will come to reuse it here and I will do imagine I will paste the address of the image and then what I will

Speaker 1: do is I'm going to reuse it here. I'm going to do Imagine. I'm going to paste the image address. And then what I'm

Speaker 0: going to do is create a little game of variables. So I'm going to say, angry, joyful, fearless, for nothing. And then I'm going to close this, reopen this, and I'm going to say, maybe, very bright environment, at the beach, full sun. And what else could we add? Another environment, at the office during meeting.

There you go. And we're going to create like that, and I have to close with a close and not a parenthesis, otherwise it won't work. So, by taking the image I found the link that was there, I added 3 options and 3 options, which means that it will give me 9 image possibilities. We'll see what it's going to give. Yes, I want to create all these variations, all these combinations.

And so, indeed, he made me 9. Why 9? Because 3 times 3. So what did that give us as a result of this story? Well, if we look a little bit, here it is, it may not be famous.

I like this image, if in reality I wonder if it's even at work or rather a police interrogation, but let's say I like this 1 pretty well as an image, I think. Here we have things that are much stranger, sometimes. I wonder how the algorithm does its job and how the thing creates that. I'm not very sure, but let's say I like this image anyway. It makes me laugh, the image here on the top right.

So here, on the other hand, it's irrecoverable, I think, because at the same time, our guy, he lost 15 years, but he kept his beard. That is to say that here, this childhood, it seems a little strange to me. It's not going to be possible, I want to get something out of it. Where it's going to be more interesting, it's here. I like it here, I think there are some good images in there.

I would say that the 1 that looks the most like our original subject is perhaps this 1. The beard was a little longer but we still keep a certain character. Here maybe not, here it's not quite the same head I think. This 1 is not bad either. So let's say we keep the 1 below here I think.

We get to this series here. So there, on the other hand, he's gone into a total delirium. And I like the image here below, somewhere. We're not in photorealism at all, but why not? Here too, I'm in...

I'm in... I'm in a perplexed state of mind. I don't know, I don't see what we can keep. This 1 again, same thing. It's a desk at the same time and a plantation at the same time.

I don't know. So, I like this image, I think there is a certain light, a certain nice thing. So here we are really very very beautiful, so this 1 here at the top right, it is just beautiful. Because, I don't know what to say, but... This is a great creation.

I don't see what we're going to do with it. Apart from laughing, it could be nice if you have someone who looks like this at work. You take the image, you recover it, you punish it on the board, it's not very nice, but... But somehow it depends on the environment. That said, this 1 is not much better.

This 1 looks like the short person in Game of Thrones. You see, it seems quite... I like this image again. I like the message it sends to you. You have all the people in the office watching you, and you're still stuck in yourself.

I think this image tells a story, there's a way to make it something interesting, in my opinion. So I put it in variations to see if we can get better. This 1 I think is pretty good, it's really nice and it's the same character as earlier, so we have something that is good. This 1 is already a little more complicated but we are maybe a little more in Fight Club mode. This 1 in my opinion corresponds quite well.

This 1 I told you what I thought about this 1, I think it's nice, but in another genre eventually. This 1 is not bad, but it's not quite the same character, it's a bit too smooth here to be exactly the same character. I find this 1 really good in a certain context to tell a certain story, I find it quite nice. So here we are, because it was the goal of the video anyway, it's to show you how to create coherent characters, characters that always remain a bit the same. If we take the character we had at the beginning and with which we started earlier, this 1, there is indeed a form of resemblance, it has moved away a bit, but it does the trick, it is probably enough compared to what we would like to tell as a story, or if it's images for social networks or something like that.

Obviously, we're not going to go with that in production mode, and we're not going to cover a magazine with that kind of image, but in the illustration of a sentence or in the way of saying things on social networks. For example, I find that this image with a little text here in the negative space here, I think there is still something to do, I think it's quite interesting. Same thing for this 1, somewhere there is probably a way around this character to find something to tell. And then there are aberrations or things that make no sense and are completely off the mark. But somewhere it's not very serious.

It's free to generate more images in a way. So why not? The thing is, how long does it take? Me clearly here, the video I'm shooting now, it's been shooting for 48 minutes, to generate images, it does take a while, and if we do nothing else during that time, we still lose a lot of time, a lot of hours, a lot of minutes in these creations. What I advise, if there is a way, is to have another computer or a tablet on the side that is made by the generation.

Yes, ok, great. And during this time, we continue to work, we continue to do other things, because otherwise we really lose a lot of time. But when I see the images that we get in total, I tell myself that it's not so bad. Here, in any case, in terms of generating a character that looks like the original character. I think that's not bad.

By playing a little more with the contrasts or with the texture, whether it's in Canva, Photoshop or the software you use for image processing, I think we can get something that is still quite interesting. The last thing I wanted to show you, beyond the character generation, the pan, the zoom, the x2, etc. Is how we get by here with our gallery when we want to select several images. So here I'm going to do a little reload so that it recovers the last images we just created. So there it is.

And so I say to myself, OK, I'm going to get all the images of my character here to make a zip and to be able to use it. How can we do that? Well, now here we have this option to select and choose the images we want to keep. So if I do that, be careful to click well outside the area that just added there. So I get my images.

And then at 1 point or another, it will simply create a zip with my images and we will do the downloader here so I will do this prepare a zip we will say here that we do not need to make a huge batch size but I have already downloaded 1050 images like that in 1 go and he created me several packets several zips of 250 images I tried with 1000 it didn't work so here it is and so I'm going to download all this he's going to put the images in a zip to retrieve the zip We have our images that have been downloaded and now we can use them in posts and things like that. So here's the result, even a little enlarged, I still find it quite nice, quite clean. I think there's a way to do something quite interesting with that on social media in particular. So on that, I think the main thing is go ahead, create, show me what you do. I want to see a little bit of your different creations.

I am very tempted by the story of Stable Diffusion and the story of creating his own model to generate images that are pretty good in the vein of what I would like to do with CreaCity. But I just don't have the time. I know that Thomas Boyle does some pretty incredible things with this story and I would be really super tempted to try to create the world of CreaCity, but thanks to Stable Diffusion and a model that I will train myself to generate images that I want to see, and not just the images that Midjoney thinks are interesting. And then, I think there will be a lot of things to do. Here I will soon go on vacation and, in my opinion, generate images.

I will generate some during the holidays, because it's the kind of thing you throw on the tablet you generate some images, you keep playing cards and then you come back and you look at it from time to time it's even funny to show the children and the parents I find, we show them a little bit what the engine has created, among other things the image of the guy with his grunt there, it was still quite disturbing. So here it is, once again my name is Vince and thank you for watching this video. Don't forget to like, comment, share, subscribe and ring the bell. Be nice in the comments because trolls like to know when they dance, sing or give hugs. See you soon.

Thank

Speaker 1: you you

See all Krea City transcripts on Youtube

Nouveautés Midjourney : Pan, Zoom, création de personnages cohérents