The Podcast
Video
Show Links
- ChatGPT - https://chat.openai.com/
- PhotoShop - https://www.adobe.com/products/photoshop
Help the Podcast
- Apply for Shopify Store Consult - https://jadepuma.com/pages/podcast-consult-application
Transcript
Hey, Scott Austin here. In this episode, I'm going to cover A.I. for Image Generation.
now, we're going to be talking about images and image generation. I'm going to have an accompanying video for this podcast. So if you listen to the audio version of this, you can later on check out the video version and see the images that we're creating today.
Now, for those of you that are counting, this is the third episode I've done on air in the past year. And that's because I'm a big believer in content creation for building e-commerce brands. And I think that is a very valuable tool that every store owner should be thinking about adding to their content creation strategies.
And in the past month, there's been some significant improvements in graphics and image generation from AI.
And I'm going to cover two of the tools that I use to do A.I. generated images. The first one is Dolly three, which is from Openai, the same people that bring us GPT. And you can now use chat GPT three to create
specifications for Dolly three to use to generate images. And it makes image generation even easier than just using an image generation alone.
But having a chat tool that's an interface for you. And the other two here to talk about is Adobe Photoshop and how they've added their own AI tools into the Adobe suite that you can use throughout your graphics.
And even though these tools just came out within the past month, both the chat chip one in the Adobe Ones, I'm now using them almost daily at my agency.
Now these tools don't give me everything I need from content generation. You still have to create your own graphics and photos when your requirements exceed what can be done through AI. But I can do a lot of the lifting for a lot of your needs.
So now I'm going to do a show and tell of each of these tools just so you get an idea of what is possible and some of the steps that you can take or best practices to get the most out of what's available. Now, the Bristol we're going to talk about is Gypsy and Dali three.
They're both from the same company, Openai. And now if you were on the pro plan of GPT, which is $20 a month, you have access to the Dali three interface. And I've got that on my screen right now and I've already gone through and run through some image generation scenarios ahead of time. And that's because when you create an image which at CBT it'll take some time, maybe up to a minute every time to do another round on your image creation.
And there's also a cap on the number of images you can create at any given time.
So I did all of this ahead of time. So we don't run into the caps and we don't have those pauses in Image Generation. So the first image I wanted to create was just a graphic. And the graphic is actually for this podcast episode.
I actually the instructions I said was, Give me a vector graphic to use as a cover for a podcast episode about generating images for your e-commerce store using AI. And every time you can actually see an image query, it gives you four images that you can choose from.
in all four of these images, you know, I asked for a graphic, so they kind of look like they were made in Illustrator or the vector tool.
They're not photorealistic. In the first one he gave me is the one I like the most.
it's actually a robot at an easel with a painting and a whole bunch of graphics around it to make you think about, you know, photo creation, image creation, e-commerce and that kind of stuff. Second image you gave me was a shopping cart, because when you talk about e-commerce, there is a AI in there because we asked about AI and what Chachi Pizza is doing.
As you scroll through these different images, it renders what it's doing is it takes the small prompt I gave it and expands that into four different props. So Chachi begins adding its own intelligence to this in giving more nuance to Dolly three on what to create. And that's how you can come up with four different graphics from one query because Chachi begins interpreting it four different ways.
third one I didn't like, it just wasn't it was confusing to me. It's like it was cityscape and clouds and rain and it didn't make sense to me. But, you know, that's okay. And then the last one definitely looks very e-commerce. Although it doesn't really say to me image creation or A.I. as much, but with the four of them, I could look at them and say, Well, I like this or that about them, and then iterate on that.
So from my second query, what I wanted to do was get the images to be more along my brand colors. So I have a green, blue and yellow brand color. So what I did is I updated the instructions with update these so that the primary color is and I gave a hex good for green and additional colors for and I gave a hex code for blue
and a hex code for yellow or gold.
And what it gave me did listen to those instructions at all. So it just doesn't understand the concept of hex codes. So I got four new images, kind of similar to the ones I had before, but a little bit different because every time you iterate on it, it's just going to give you a new run. But it didn't incorporate the colors I was looking for.
So then what I did for my third query is I said, make the primary colors of the graphics Jade green and each one of the four graphics is predominantly green, but they're not the same shade of green. So you can do instructions and it can choose to listen to it or not listen to it. And sometimes it's kind of interesting on what works and what doesn't.
But now I've got four images.
the first one of the robot painting again. And the second one is like this mind hive type of thing. It's also in green. The third one still has that city skyline with clouds and rain. That makes no sense, but it's also green. And the last one has, you know, green on the very e-commerce looking image of a monitor.
And sometimes I really messes up on the text right on this one. This one I'm looking at now says I enhanced e-commerce, but the Age has three vertical parties instead of two. And commerce is like instead of two M's it's at m and and and two is it's kind of wacky but it's all work up better than it used to be in AI generated images were most of the time the text is making a lot more sense than it used to.
So now I've got these four mainly green images, and that's more along my brand. So the next thing I did is because it gave me square images and I want to use this as the blog header on my Web site for this podcast episode. And my blog images are all for the three aspect ratio. So I said, provide the image in a 4 to 3, you know, four colon three aspect ratio.
And I tried to do this before. I
now provide images that are more landscape and it totally added landscapes to me, to the images instead of making them landscape in size, right? Or more horizontal. So I found that doing the four cool and three aspect ratio works really well.
And now what I have is the same four images. But instead of being square, they have a 4 to 3 aspect ratio. So we're definitely starting to hone in on something that I can use on my blog post. So then I really like the robot one every single time the robot painting on an easel. So for the last three, what I did is I said, Give me four variations on the iRobot graphic.
So what it gave me is four robots with the predominant color and green in that 4 to 3 landscape aspect ratio. On that said the robot, it gave me on the iteration before I liked that one better, but now I have these five different robots painting on an easel concept that I can pick from that use the right colors and the right size for my needs.
So in a few queries, I was able to generate a image that I'm actually going to use as the cover image for the blog post. For this podcast episode where I used to do in the past was I go to stock photography or, you know, make my own image and I'm not a graphic artist. So they never look at and I think this is a way to quickly and easily get the point across that, hey, this episode's about AI in images.
And I think that you could do the same in a lot of your graphics. Now, if you're going to be doing like an infographic, we actually want numbers and stats and exactly precise things in there is not going to do that for you yet, but you can totally use AI for graphics. It set a tone on what you're trying to do.
So that's a graphic. Let's go to another example inside of Chat GPT and then this one here. I have a client who has charging solutions, so like portable batteries for charging your phone and your laptop. And that got me this idea. And they have all these industry vertical pages were building like for sports, for education, for outdoors. And we're actually having a challenge coming up with images for each one of these industries that we're targeting and all the marketing that we're putting together.
So I asked Jackie Beatty to do was give me a photo realistic image of campers by a campfire in the mountains with a lake in the background with a battery for charging laptops beside them. And what I got was four images of exactly what we're talking about, but four different takes on it. And some of them are a little bit better than the others.
And one of them, for whatever reason, there's like a Ferrari car, sports car by the lake, which makes no sense, and a camping thing. And some of these laptop battery units are absolutely massive, but each one is a little bit different. There's there's one, for example, where the laptop battery image is in the forefront, new focus. And the campers are in the background and out of focus.
And that's really nicely done. So in looking at these though, it gave me what I asked for, which was, you know, campers, plural. So I said, this has to be like it's like 12 people sit around a campfire, sometimes even more than that. So then the next thing I said, make the campers a couple. And that's the only additional information I gave it.
And when it gave me was four images of a couple. And each in this situation, each couple was a man and a woman. Sometimes it'll actually give you two men or two women, so it understands a little bit about diversity. But also in this they got four images. I got eight people, but two of those images, the people are facing away from me.
So then what I said is make the campers face the camera. So now I've got four images and in the first one I've got a white couple. In the second one, I've got a white couple. In the third one, I've got a white couple. And in the fourth one I've got a mixed race couple. And this one that actually says in the prop that she actually created a photo of the diverse couple on the other photos.
It doesn't say anything about the ethnicity or diversity of the people. So in looking at that, what I wanted to do and they're all based in the camera now, which is nice. And if you look at the photo realism moment, none of these are, you know, photo realistic as in I would mistake them for a photo close up, but a few of them are photo realistic enough that in a small enough size or from far enough away you would actually think they're a photo.
A couple of them are a little more graphics looking and you can totally tell it's not photo realistic but a couple of for you on first blush which I was absolutely amazed at this level of quality of you know from a realistic in air quotes imagery because this was not available just a few months ago seven in looking at this, I was thinking the battery packs were too big.
So I said, Make the battery black because my client's battery packs are all black and a bit smaller. So the next round it gave me all black battery cases, which is nice. And in the Bruce one it gave me two men, even though the directions say photo of a male and female camper of diverse descents. Right. I got two white men.
All right. So would still be fine. But the point is, you are not in control of exactly what the AI is going to give you. You can give it directions. It's going to interpret it however it wants to. You're going to get directionally the way you want to go, but it's hard to control precision what it is that you get in the second bird.
I've got a man or a woman in a black case there photo. I've got a man or woman in a black case. And in the fourth photo I've got the man and woman with a little ethnic diversity and a black case.
So and the next direction, what I said is diversify the people. Right. And if I look at that one, my first image is is a diverse couple, diverse and ethnicity.
The second one, white and maybe Hispanic, or it actually serves Hispanic female. You know, she does look a bit Hispanic. So that's good. The fourth one I got a white camper and a female is of indigenous descent. I can kind of see that. And the last one is interesting because it says a photo of a male camper of South Asian descent.
And he told a white and a female camper of East African descent into me, she looks more Asian. So like I said, you can give it very specific directions and it may or may not listen to you when it gives you back something.
So then as I was looking at that, I'm like, I got a little more diversity to pick from.
And I said, All right, let's make the people 55 years old. So now I got four images and these people do look 55, maybe even, you know, a little bit older than that. And the first one I've got, you know, a diverse couple in ethnicity. The second one looks very white and maybe Hispanic for the woman. The third one is diverse.
It looks more cartoony, though it's less photo realistic and which it has been all the way along. And in the last one, we've got a diverse couple. But but the adding Ange to them is totally realistic and kind of awesome right You compare the the photos from before and the only thing I said was hey now make them 55 years old and you're like, boom, It just does it.
It's really nicely done in that, you know, for them. An example, I now have a ton of different photos that I can pick from and you know, use whatever ones I want in whatever situations I want. And I might have to go into Photoshop and take out the the battery it added and actually add my client's battery in.
But that that's totally possible. But now I've got all these backgrounds that I can use in the situation that I'm looking for. So now I'm on to another scenario that I did and that is
I have a client who has coconut substrate that he sells for the bottom of your cage, for your reptiles, like your snakes and your frogs.
So I just ask for a symbol, give me a realistic throw it over ball python, which is a type of snake on coconut substrate. And I was blown away when, you know, the photo realistic quality of these you know, I think it's harder to do a photo realistic reunion than it is to do a photo realistic snake. Now if you look at these, you can still tell they're computer generated.
They're not that photorealistic, but the textures and the lighting and the shadowing, it's just unreal. I would blown away the first time I did this and what it did is and the first one, I've got the snake. And he looks incredible in his our panel of coconuts is actually not, you know, chopped up coconut substrate, which is fine.
Right. Same thing with the second one. It's like the same image from a different angle. Same thing with the third one and the fourth one. They all look a little bit different, but it gives you a place to start. And like I said, the photorealism of those is amazing. Now, my client actually has a logo of a specific type of ball python and it's a blue eye, loose cystic ball python, which is basically a white snake with blue eyes.
So what I said is give me a realistic photo of a blue eye, the cystic ball python wrapped around a coconut. Right. Because they're little, but we actually have a whole coconut and the snake wrapped around it. And that was done in Photoshop. Now, if I look at these images once again, I was absolutely stunned by the quality and how realistic they looked.
But in every single one of them, the ball python is wrapped inside his is in a circle inside of a half cut open coconut. Right. And I wanted it wrapped around a coconut. So then what I said is, all right, the snake should be wrapped around a whole coconut and what he gave me back was even better looking images of a snake in a coconut.
But none of them was the coconut whole. It's chopped in half, and the snake is still inside of the coconut. So on this one here, I realize there's probably going to have to spend a little more time and figuring out how to build up and get Dolly to do what I want it to do in this situation. But the quality graphics just totally blew me away.
So now we moved on from that. I'm another client who has meats that they sell that they raise in the mountains of New Mexico, right? So I said, Give me a photo. Realistic image of a herd of yaks. Yaks is one of the types of meat they have grazing in the New Mexico mountains. It's one to see what it would give me.
And it gave me, you know, totally what I asked for. You know, there's just three or four different versions of yaks grazing on grass with mountains in the background, slightly different terrains in each one. The mountain peaks are all really steep and sharp, which, you know, maybe we can edit on that. But it's just amazing to see how it could do like, it's a whole scenic painting in just a couple seconds time.
And then I moved on from that and said, Be on my work site for Jade Puma for my agency. One of my branding elements are different Stock photos that I have of store owners in front of their, you know, in front of their wear. So inside of their shop, you know, just look into the camera like happy, confident people kind of thing.
You know, it can probably understand my branding is if you work with Puma, this is going to be you you're going to be happy, successful e-commerce person. So I said, give me a photo with this image of a happy owner of a shop in front of their wares. And what I found is if you ask for a happy with Dali three, it is scary how happy these people are.
Elham Another example in a little bit, but I've got four store owners and the beautiful thing here is I didn't ask for anything in diversity. And what the Church Committee did was on the first image, it says, Hey, give me a middle aged Asian woman. On the second one, it said Young African man. The third one, it's a Hispanic woman, and the fourth one was an elderly Caucasian man.
So, Chad, Gypsy is automatically asking Dali three to give diverse images in each of the images. I've got this person standing in front of their shop and in a couple of the the smiles for them happy that their smiles are just a little bit too big kind of thing. But still absolutely fantastic because you can do something, you know, you might be more specific with what types of merchandise.
So let's say you were targeting on an e-commerce agency targeting just bicycle shops. You could ask it for, you know, five or six different people in their bicycle shop and you could show that diversity of bike shops, but consistency that they're all bike shops kind of thing.
So the next thing I said is make the images wider so they can be used as banner images.
And it gave me basically no change, right? So yeah, this is what I was learning. I got to tell it, I want a specific aspect ratio for it to, you know, do that because just say naked wider didn't change anything.
And then I even went extreme. I said make the images three times as wide as they are tall.
And it just didn't listen to me. But when it did that, it did some really weird things on focus. Like, like the image I'm looking at right now is of the man and he's
it's stretched out into the depth. There's a lot more depth to these. So somehow by seeing I wanted, you know, three times wider, it gave more depth to the images and a different view on it than it had done before.
So I think it's one of the things just you keep playing with it, you'll see the cause and effect of what you ask for and what it gives you. So then I do another client who is big on Easter candy and we're building out their Easter candy landing page. And I said, Give me a photo. Realistic image of kids on Easter with a happy family.
And this one here, Happy is just way, way too happy. And the first image, there's a family and the three guys. So there's three women, three guys and four kids. The three guys look like triplets, right? And everybody is super white in this photo. And they're very, very happy. In the second one, I now have three women, two men and four kids.
And, you know, they're also really kind of happy. But, you know, it's this really there's a lot of detail, all because it's amazing how much detail this one's taken indoors. There's Easter decorations all over the place and rabbits and eggs and all that kind of stuff. In the third one, this was taken outside. You know, there's tons of people on this image.
Also, ironically, the images of someone taking a photo of the scene. Right. And the camera, the image in the camera is not the image you see. So, you know, that little level of detail, they were kind of missing. But there's a ton of detail on this one also. That's just absolutely extraordinarily amazing. And the last one, it's let me just give you the text here.
So what she achieved, she came up with for the project is realistic photograph capturing the essence of Easter with the family, children laughing and running with Easter baskets, parents sharing a moment and a backdrop of spring flowers and bright skies. And it's just absolutely over the top with detail. But interestingly, on this one, there's a two parts man one And there's three kids, two boys and a girl, the girls in dad's arms.
There's a whole bunch of people in the background also. But in the forefront are these five people. But the mom is totally messed up by a guy. So the we can see the full mom, you know, her whole body. She's only got one leg and maybe the other names behind her because all running, but in between the full mom and the full dad underneath the girl, there's another mom's half body with two legs.
So, you know, this imagery totally unusable because somehow they just totally messed up. And that does happen, you know, in are the human bodies because they're really hard. And like I've seen before with the snake images that I was looking at, the snakes were just extraordinarily detailed and accurate. And part of that is because, you know, a snake's body is not as complex as a human.
And the other part of it is we all know human bodies better than snake bodies. So give me their mouth. You know, the snake's mouth was inaccurate, but I don't know that because I don't know snakes, but I can look at this one to go, Well, that one was only got one leg. That's a mistake. Right. And you just eerie when you look at it.
So then after I got all these really happy people, I said, Can you make them a little less happy? And the first image I got there a lot less happy. In fact, they're almost sad. No, their hands are down. It's it's so you have a gentle in the directions that you give it.
and the instructions for the image say photo realistic image of a diverse family gathered in a garden during Easter with kids searching for Easter eggs with mild enthusiasm, parents observing calmly in a modestly decorated Easter basket nearby
so you could easily tweak those words of mild enthusiasm and parents observing calmly
to increase their their excitement a little bit.
In the second image, which is the indoor one, I would say. And there was a lot of detail before. And every time you do a new image,
but it's a redraw of what the previous one was.
So it doesn't start from the previous one. It just starts with the same instructions as the previous one. But ends up with a different result because it's redrawing it every time. And in this one I've now got what looks like four grandparents were before there was also four, but one is man and woman is a couple and the other two are two men as a couple.
And the interesting over two men and a couple, they're sitting very oddly like one is almost in the lap of the other. And between the two of them there's only three legs. So I'd lost a leg somewhere along the way, resting. Even so, there's a lot going on in this image, just like I'm the last one.
And then one of the other images that gave me the photo realism on this one was much improved over the other ones.
There's even in the foreground that people are in focus in the background, they're less in focus, and they even put a little bunny in the Easter basket and it looks like a real rabbit, you know,
And the one where they were running through the woods all crazy or running through the fields, all crazy with mom having, you know, one leg, that image looks a lot better.
But there's also a person in the background, you know, and he looks to be like a 12 year old boy. And he's holding two six year old girls, one stacked on the other. So it's not a realistic pose. They're really nicely done. This was a little bit more surreal. But if that's the tone you're going for, that's great.
And what you can we do in Chatty Betty. And you saw me do this on the Robot, One of the Jade Puma graphic before is, if you like, one of the images. You can describe which one you like and say, you know, let's focus on this one and do this with it, and it'll give you four versions on that.
And then for the next image, I was actually doing a Halloween email campaign and all I said was, give me a for a realistic image for Halloween. And it gave me four really awesome images. And they are very Halloween. There's no mistaking this setting the tone that this is a Halloween email campaign. And I actually used one of the images I just in one query was totally happy with what I got used one of the images and put it in the email.
So, you know, that took me like 5 minutes or less from during the query to to jets empty to actually adding to the image to clear the you in the email campaign that I was building, which is much less time the demon go look in stock photography sites for what you're looking for.
In the last example I want to give with Jack Beatty and Dolly three is for making of a logo.
So right now I'm developing a marketing course for new Shopify store owners, and in that course I'm using an example of Yo-Yo Wizard as my demo company as we build out the branding for the course. So what I said to Charity, Betty, is give me a logo for Yo-Yo Wizard, an e-commerce store that sells use. And it gave me three of them, right?
And each one has like a wizard hat. Not all of them have a wizard. Have most of a wizard hat or a wizard silhouette and a Yo-Yo. And you know, it's a place to start. So then I said, Make them more landscapes. They're trying to get wider images. And in this case, it actually did make them more landscape.
But what are the images, which with a logo before is now a logo on a card in somebody's hand, which you know, I don't need that. It's a flat logo kind of thing. I mean, what I realized as I started go through these examples is because you have so little control over the consistency of things, it's really hard to just say, now tweak this and tweak that.
So I said, you know, the one that I liked, I said, Let's go deeper with the one with the wizard hat on the Yo-Yo. So then it gave me four versions of that. So I got four different look wizard hats with yo yos or things that sort of look like yo yos around them. But then it just became really hard from there to on to zone in on, on what I really wanted.
And then I said, Let's simplify it a little bit. Is there a lot of detail in them? I don't think logos should have a lot of detail. So then it's simplified them a little too much. And I said, Not that simple, and it simplified them a little less, but for some reason it was given them all an outdoor theme.
And I think that's because I said the word landscape. I wanted the landscape as in an aspect ratio or size of the image, but it was interpreting that as, let's throw in some outdoor imagery in them. So I said, Get rid of the outdoors rivers and trees concept, and it totally did not listen to me. It gave me new versions and they were still outdoors and trees and rivers.
And I said, So they still have outdoor elements and it still did not listen to me. So in that case, I would probably start all over again with the new context. There's no history in church that it's leaning upon, so that is some really good examples of using church, Betty and Dolly three to make images. I think for things like blog, article, headers, navigation elements, emails, you know, you can do a lot with graphics and semi photorealistic images from these tools if you're doing, you know, product photography, navigational elements with, you know, big banners, you're probably still going to be doing that.
The traditional we've always done it, but you can start use it. I for these other lower resolution and much more higher frequency needs for creation of content than you have in the past.
So now let's switch gears away from Jeopardy and talk about A.I. inside of Photoshop.
So Photoshop has been putting A.I. inside of their tools for a little bit of time now under beta. And now that's finally launched and it's just amazing what you can do. So I'm going to show in a bunch of stock Pro Tools that I downloaded from Unsplash. So the concepts of things you can do and things that I have been doing regularly now, and the beautiful thing with the AI in Photoshop is you're starting with your photo, so you're not asking me to generate a scene and to come up with ideas.
You're going to start with your photo and then you're going to have the AI in Photoshop improve upon that photo for your needs. And one of the first things that I use this for is in today's world, we have a lot of portrait photos taken by phones, and I like my imagery in a lot of situations to be square or doing product photos or collection photos or for banners.
We might also make the more landscape. So it's really hard to take a portrait photo and use it in a landscape manner. And my favorite feature right now in AI on Photoshop is to create a banner image from a portrait one. So on my screen or now a prototype, I've got this landscape image at this mountain with a house in front of it
All I'm going to do is I'm going to increase the cropping. So I'm going to make the the image wider and I'm going to tell Photoshop to generate an expanded image from this.
And then it could be okay, but it's going to take a second to do it. It's thinking and now it's generating.
Now the results are in and just like in chats, CPQ, it gives me four options for me to choose from.
What Photoshop is doing is give me three options. So here's the default one. And I have this mountain and I expanded it and it just took the slopes of it going on both sides and just extended it out. It's just amazing. If I look at the second options, ready to look at my options, you go, Well, which one do I like the best?
The second one I like a little bit more because off to the left there's a little less fog. So I can see more of the details. And if I look at the third one, I don't like it as much. I like the second one for whatever reason, but I get to pick the one I like and then you can go back and change if I want to change my mind any and this is what Photoshop does, it's called a nondestructive process.
So right now I don't like any of these. I can just delete the layer because it adds a layer for each of the ones that you get to pick from. And I can turn that layer off, see what the original photo looked like. I could turn that layer back on and look at the improvements and judge whether I like what the air has done or not.
But that's not even close to all you can do inside of Photoshop. So in this image, I've got a house and all I'm going to do is I'm going to select near that house and I make a selection box and it brings up a menu item called Generative Fill. And I'm going to say that I want a log cabin instead.
So I've drawn a box around the house and I'm saying, Let's change this to a log cabin. And the air is going to generate three options for me to choose from.
So now that it's done, I can see this, this really rustic looking log cabin I got three options and they all look pretty rustic. So maybe what I want to do is instead of it being a log cabin, maybe I want it to be a modern glass house.
Now, as it's generating my three new options, it's got to keep my three log cabin options in this layer. This layer is now going to give me six options that I can choose from, or I can just get rid of that layer and take none of them and keep the original house that was there. So there's glass house, number one, two and three.
I like three the best, but maybe we want to go with a
sleek
concrete
house
with
large
windows.
now it's going to generate the sleek concrete house with large windows. Give me three options for that. So now I'm going have a total of nine options that I can look at and decide what house I want to be there.
I see the three options. I'm actually liking that one, maybe that one. I'm still going back to the glass house, one that I had before. I'm going to leave it at that so you can take an object and replace it. You can also delete objects. So next to my house I have a grove of trees. So until I draw a box around the grove of trees
and I'm going to just hit the delete button,
had this group of trees next to the house that I want to get rid of.
So it's going to select the drop box around the trees
and I'm going to hit the delete button
and I wanted to do a content aware deletion. So what it's going to do, it's going to look at the content around the image, right? What the textures, the colors and all that kind of stuff and try to fill in the best it can.
And sometimes this doesn't work well when you've got an image right there of the house, and that's what you see right there is is not working that well.
I'm going to undo that one because it didn't work that well and that's because there's a house right next to it. So now when you go to play with a little bit in Photoshop, but is another fun one that than I've done in it.
So now I've got this the steel in the background and I'm just going to put a box around that and I don't even have to use the box, or I can use the lasso or tool if I want to, and go along the ridge line of the the field instead of it being a square box. And I can say
add
sheep
grazing
and it's going to add some sheep to the hillside for me and I'll be able to look at it.
And you know, I've three options and I can pick the one that I like and keep that one.
So now I have three different sheep to choose from. I've got this one. I don't like it. It looks cut off on one side. The second one, that one looks a little bit better, but I've got some issues in it also.
And the third one looks a little bit better. It's got smaller sheep, which is more realistic given the distance away that I pick for the fields. I might keep that one
go on to another image
and in this image I've got a green sneaker on a background of green leafy shrubs, it looks like.
So what I'm going to do a Photoshop verse is I'm going to pick the
object selection tool and just highlight the sneaker. It's automatically going to select the sneaker. Even though the sneakers got a lot of green, it got a camel Look, it can distinguish between what's the sneaker and what's the shrubbery background behind it. So now it's automatically selected that I'm going to hit delete and we're going to do the content aware fill.
Sometimes it works really well, sometimes it doesn't. It all depends on, you know, how much information the image has. So there we go. And it's left to the shrubs that, you know, it's added to where the sneaker was a little bit darker than the other part of the shrubbery is. But I turn off the. Yeah, you really can't tell that, though, when you turn off the selection tool.
And then what I'm going to do is I'm going to add a new object by a selection box. And I'm just going to say tennis ball,
and it gives me three different tennis balls to pick from injuries to me, one of them know one of them has some leaves and kind of Friday give me, you know, more realistic look on the highlights of or a little too high but I've got three options to choose from there and it's got a nice next thing going to do is I'm going to draw another box.
And what I've learned is the size of the box she draws important to make sure that the image that's being added isn't too big or too small for the scene that you're putting it into. On on this one here, I'm just going to say football
now. Adobe is an American company is probably going to mean American football versus a what I would call a soccer ball.
We'll see.
And I've got a pretty good looking football there. They give me a brown one to give you a white one and another brown one. Well, that's nice. That is really nice. I'm just totally impressed of this. So let's move on to another image, and I'm just trying to give you ideas of what you can do inside of Photoshop tools and where the limits are and a little bit of tips and tricks that I've learned.
So on this one here, I've got a young man sitting on the steps outdoors and this is a portrait image. And let's just say I want to make it square or I want to make it, you know, even more landscape than that. So what I'm going to do is in the crop tool, I'm going to add some wet to both sides.
So it's about what I think is square and we're just going to do a generative expand and, you know, the steps that he's either at kind of an angle. So I want to see how good it does with the angle of the steps.
that is stunning. I had done this one before and the steps did not stay as straight as they do on this one.
This is just unbelievably good quality. I can't tell at all in looking at it right now where the added parts started, you know, sometimes you see a little by the seam. This just looks really, really good.
So in my next image I have mountains, silhouette of a mountain with golden sunset sky behind it. And all that to do is just be a little bit silly and say, I want a UFO in the sky
and my first UFO doesn't look that good.
Second one's better because it it takes the glow from the sky and the third one is very silhouetted also, but not bad. And I can now generate three more to see if there's a better options that it gives me.
And now I've got six options to choose from. And Neil, they're all they've got a similar concept the viewer.
So you've always got a locked in one thing but there's differences in colors and light and this is the one I like and I would keep that one. And we're going to move on to another image. And in this one, once again, I have a mountain that's kind of silhouetted and I got a blue sky behind this one.
And in this one I just feel like I want to see the moon rising. So I'm going to just draw a square box over by the mountain and I'm just going to say Moon.
My first moon does not look good. A second moon looks awesome. It looks like our moon where did. Is it? Give me three out of four millions, not Earth's moon.
The second one is Earth's moon. And that looks really good. Now, I wonder what would happen if I said, Well, instead of the moon, I want Jupiter,
because maybe we're on a moment of Jupiter when this photo was taken after
it gave me some weird shark looking spaceship size to be the Jupiter from
Lost in Space. It gave me a burn.
And the third one is sort of like Jupiter, but with Saturn rings in a couple moons. So we'd have to work on that one a little bit. And then just to show you on the expanding the size of the image, because we have this portrait image, let's take this one and make it all landscape also and see what it gives us,
these generative,
these generative expands on images are just absolutely amazing.
And this is the one that I use the most and stuff that we do take our portrait photos and making them square a landscape depending on the use that we're trying to get out of them. This one here is just another example of that. I got three different versions. They all look pretty good.
So my next image is it looks like a state capitol building and the sky in this one is really white or gray and hazy.
So what I'm going to do for this one is I'm going to go to the selection menu. It just sky and I'm going to let a dummy say I pick what it thinks the sky is.
And in this image, there's a state capitol and trees on both sides of the photo. So it's picked the sky. Looks like it's done a pretty good job.
And what I'm going to do for generative fill is I'm just going to say blue sky,
because this image with the current sky is really bland and boring. It's not bright, it's not happy. I want to see if we can use AI to take This photo was the only one I've got and make it a little more appealing.
So in version number one, I still have clouds, but I can see some blue in that sky. It's a bit brighter and I think it did that because around the trees it's hard to fill in all the blue. Not really sure. Second one, no improvement at all. It actually gave me more branches out of the trees. It said it was absolutely awful the more I look at it.
And the third one, I have a blue skyscraper, I think behind it. So only one did it give me a blue sky. So we're going to do I'm going to say blue sunny sky. So if that helps. Similarly, give me three more versions or the first one to give me with the blue is a lot better than what was there before.
a little brighter of a day in that one.
Still, I would not use that image, you know, in a high resolution on like a banner, but you could totally use it in a smaller thumbnail or in an email.
But in Sunny Sky one took the top of my state capitol dome off, but the sky part of it looks good.
Ten It looks good. Also
in three, it's adding some foggy rooftop to the back of the state capital that wasn't there before. So this is a complex picture with the trees and the state capital. So it's trying, but you'd have to play a little bit to get it to be good enough to use.
Now, the last image I have is just of a beach and some rocky cliffs near the beach and an ocean.
So the first thing I want to do here is just draw a box and let's put on a sailboat because the ocean's got a blank right now. So I'm just going to say sailboat.
And I've got three versions of a sailboat. First, what is a bad second one is not bad. And then the third one is not bad.
So those are all nicely done. I'm going to draw another box on the cliffs and I'm just going to say
ia2
hiker's
is to see how we can add content in different areas.
so my first image to hikers looks really good. My second image, I've got three hikers and they're all wearing bright red to look like they're in a hiking team and gave me three, not two.
And my last one is a little bit too blurry. But of those, one of them is absolutely fantastic.
so now what I've done is I've added a couple elements to this photo, but I can also delete elements. I've got a swimmer in the water that I wanted to get rid of. So what I'm going to do is I'm going to grab the object selection tool and see if I can pick that person.
Sometimes when it's that small, you can't, but it looks like I can. I'm going to delete it with a content aware deletion, so it should just take that person out. And if I show the full screen, they're yeah, they're gone. That looks nice. We could also, you know, get rid of this person on the beach. Let's say
in my show that full screen and this one here, you can see there's a little bit of an outline around it.
So if that happens, sometimes you just draw a box around that whole thing or the lasso and just have it do a content aware fill of that area again and see if that. And that helps a lot.
So you can see there in this image with added a couple elements that were you know kind of lacking life and we where it's too crowded.
We deleted a couple of elements also and this is how you can do the same thing with your photography. Let's see, you've got a great outdoor photo shoot that you're doing. But there was some annoying person in the background in a clown suit and that and you didn't want that your photos at all. You could go into Photoshop and just take that person out and keep the great photo that you spent so much time and energy creating.
So that is an overview of what you can do with AI and image creation in editing in today's world. Right? And this is only going to continue to get better. All the stuff I've shown you I could not have done, you know, six months ago, a year ago, because the tools weren't available, but now they are now both.
Photoshop is a paid service and I'm on S.A.T. Pro, which is a paid service. So you're not getting these tools for free, but they are, I think, worth the money if you're going to be using them quite often.
So go ahead and start using these in your content creation, especially for your emails and your blog.
And as they get better, I think we're quickly going to be amazed in just a year or two years time what we can do with these tools, like what I'm looking forward to, is taking your transparent background photo of your product, uploading that and see now I'm getting this to me in these different contexts with this type of background
and that type of background, I don't think we're that far off from stuff that, you know, 20 years ago it was science fiction, but it's going to be our near future Pretty soon to try and tools out. And thanks for listening.