How to Use DALL·E 3 in ChatGPT to Create Unique Images

You can now generate beautiful AI images and this right inside chat GPT

That is, if you are a plus user for now. And this without having to come up with complicated prompts or even figuring out how a Discord server works

Image journey. The tech is called DAL-E3 and it's by far the easiest way to generate AI images

In this video, I'm going to show you everything you need to know to get started with DALI3

Let's go. What is up everyone Ronnie here? Welcome back to our channel. This is the right place if you

want to learn about Canva, AI, chat GPT, image generation. We cover all of these topics right here

on the channel. Today we are going to take a deep dive into OpenAI's DALE3, which is the

image generator that is integrated with chat GPT. I'm really going to cover every single

aspect of DALE3. But before we do so, you need to know one important thing. My goal, my goal is

for this video is to bring you up to speed and show you how easy it is to generate images with

Dali 3. So let's jump into my presentation and get started. The first thing we need to talk about

is who can actually access DALE3. As of today, the date I'm shooting this tutorial, DALE3 is

only available as a feature for chat GPT plus members. So if you're not a chat GPT paid member

you will not be able to access DAL-E3, at least not as of today

So be warned that in order to be able to follow along with me in this tutorial

you will need a paid chat GPD account or chat GPD plus account

Next, let's talk about how it works and why it's a little revolution in the world of

AI image generation. So how it works is pretty simple. We are here in a classic chat GPD

conversation and so I am using common language. For example, here I said, hi, can you help me generate images in here

To which ChatjPT responded. Of course, please provide a detailed description of the image

you'd like to generate and I'll create it for you. The more specific and detailed you are

the better results will be. Now, that seems like what any text to image generator would

tell you. But there is a difference here. The difference is that we have our good old friend

ChatjPT here with us to help us in our journey to create beautiful images. So if we look

the Dal-E landing page right here on OpenAIs website, it says DAL-E is built natively on chat

GPD, which lets you use chat GPD as a brainstorming partner and refiner of your prompts

So just ask chat GPT what you want to see in anything from a simple sentence to a detailed

paragraph. So we could indeed go the difficult way and write an elaborated prompt like we would

in mid-journey, for example. But there's also a shortcut, and that is to rely on

rely on chat GPT's capacities to help us refine the prompt from a simple prompt

So let's go ahead and try that. I'm going to paste a simple prompt right here in the text box

It goes like this. Can you show me some examples of sleek and modern app UI for an ice cream

delivery service and UI stands for user interface right here. So this prompt is relatively basic

So I'm going to run this prompt and you will see that chat GPT will be calling Dali 3. So it's kind of like

like how plugins work, but it's just calling and relying on Dali 3 for creating some images

Okay, so let's see what it comes up with. And I'm not going to speed up the video so you see the actual speed at which it's been created

And there we go. We have two different UIs right here, one on a phone and another one on a tablet

Looks like an iPad mini. A little bit of text. So here are two examples of sleek and modern app

UI designs for an ice cream delivery service. So there you go. You have both

of your images but what is interesting here is that chat gpt gives me a little bit of extra

information for both of these images but also if I click on one I can see the entire

prompt so I went from a very basic simple prompt to a more elaborated prompt but chat

gpt did that refinement work for me that adding the detail work for me so this is where

the magic lies of using Dalli integrate with chat gpte is that we are leveraging the power of chat gpte to write and refine our

prompts so let me read the entire prompt here photo of a sleek and modern app UI on a

smartphone the main screen showcases a variety of ice cream flavored in a grid layout

each ice cream image has a brief description and a price below it the top of the screen has a

search bar and the bottom has navigation icons for home cart profile and offers so there you go

What I can see here is that every single detail, every single part of this prompt has been respected by Dali

So usually when you write elaborated prompts like that on mid-jurney or other image generator

parts of your prompt tend to be neglected or just not taken into consideration

But here, and I believe this is something Open AI has been working really hard on

every single part of the prompt has been considered in generating the image

And so I see that even dealing with text has been relatively well done

Okay, so still some little mistakes here and there. But I mean, this gives me a pretty decent mock-up

I have five fingers, not seven, and this looks sleek indeed. Let's have a look at the second one

And we'll see that the prompt is different. Okay, so we are having the UI on the tablet

The screen displays a carousel this time of featured ice cream flavors with enticing images, etc., etc

So I have my two different UIs here and they look great

I mean, every single thing looks pretty spot on if you ask me

All right, so that first prompt was pretty convincing to me. And I can see that from there onward, I can say goodbye to complicated prompts, like very detailed prompts because chat GPT is going to do that for me

It doesn't mean that I cannot do it anymore. I can and I show you that later in the tutorial

And the second thing I can kiss goodbye to is to go to to Discord to start using the best prompt, which up to today, or at least in my opinion

was Mid Journey. So bye-bye Discord server and Mid Journey, but also buy to writing elaborated

prompt. And welcome to the era of having a conversation about your images with ChatGPT

Now for the rest of the tutorial, I would like to show you a series of prompts that I've been

working with. They're not all my prompts. Those are prompts that I've discovered, that I have read on

some specialized blogs and I tried them and I thought oh this is a pretty cool

result that I'm getting here so I wanted to share them with you guys also if you

are a member of our channel meaning if you are part of our paid membership via the

join button right here go check out the community feed because we have shared a post

in which we give you the link to a Canva document in which you will have access to all

of the prompts I will be showing you today plus our little bonus with the 50 different

artistic styles to try in your prompt. Again, this is an exclusive perk if you are a

member of our channel. Now let me show you some prompts that have been working well for me

The first one is the idea of a mood board. Okay. So if you ask chat GPT to write a prompt

for a mood board for specific topic, you will get some pretty cool results. So the first one I have

here is a mood board dedicated to the world of knitting and crocheting. So this is the result. So this is the

result I got back from Dali and I'm going to click on this image so you can admire it and you

see this very cool set of instruments set of utensils that you actually need for knitting and crocheting and you can also read the entire prompt right here which has been refined by Chat Chipit

Illustration of a mood board dedicated to the world of knitting and crocheting

So you can pretty much take the prompt I gave you and simply replace what's in the brackets

for anything you like. For example, I tried with Japanese cooking instead of crocheting and knitting

And these are the two results I have been able to generate. So this one, which is more like a realistic photo

And then I have this one, which is more like an illustration. But both of them are quite gorgeous

And I really like how they are featured as a flat lay as a set of tools

a set of utensils for a specific topic. So really like that

Try that prompt, the mood board prompt. The next prompt I want to share with you is the website or app UI prompt

And here's the prompt. Could you recommend a few website UI, so user interfaces that are both colorful and easy to scroll for a business that sells emoji shaped bag

So again, you can replace anything really that is in between these square brackets

Like here instead of website, I could have said app. Instead of colorful and easy to scroll, I could have said sleek and minimalistic or sleek and modern or vintage or whatever really I want as an adjective for describing this website

And for a business that sells emoji shaped bags could be anything cars plants iPhone cases it really depends on whatever you want to design a website

or an app for so that was my prompt now let me show you the results so this is the first one

this looks like the landing page the home page of my website so again text has been dealt with in

a not so great manner but probably better than many other text to image generators and here is my

full prompt so illustration of a colorful and easy-to-scroll website you i design

for selling emoji shaped bags. So that's kind of like my original prompt

but then the details come afterward. The design showcases vibrant product images

playful typography, and an intuitive scrolling mechanism to highlight the unique bags

So we can see that illustrated here. I really love the background here of the header

which is pretty cool. Nothing is like out of proportion. Nothing is like looking weird with too many limbs

or too many fingers or too many anything. Here, the pencils, everything looks great pretty much

So that is like, was the first one that was generated. I got two different images. So this is the next one. This looks more like a thank you page or confirmation page when you have made a purchase, for example. Obviously, text is not great, but I love the aesthetic of this. And it's also very consistent with my first page right here. So again, you can read the whole prompt if you want to. Something else I didn't mention so far is that you can easily download these images with the little download button right here. So you can just

click here and the image will be downloaded to your download. So this is the downloaded images

in the PNG format. So yeah, you can easily download any of the produced images. You can also

download them straight from the prompt page right here. So with this little button right here

All right, let's move on to the next prompt. This next prompt will help you generate hyper

realistic images or photos and it goes like this. A hyper realistic photo of a pufferfish. For example

So there you go. This is my very simple prompt and these are the two images that ChatGPT generated for me

I particularly like these ones. If you take a closer look at this image, it's incredible

Like the level of details, the realism of this fish, this puffer fish, like the membrane here, the texture, the texture of the skin, the spikes, the eyes that are shining in the right direction

And it just looks amazingly realistic in my opinion. You can read the entire prong right here

Also, a cool thing about this is that you can copy the prompt. So if I want to tweak this further, maybe change the color, the direction the fish is like swimming or whatever detail

I can just copy this prompt, paste it back into chat GPT and just add a few lines of detail or extra

instructions for this image to be completely transformed. I will show you that in a second

But for now, just know that you can copy easily, copy your prompt from here

And the second image, this one with a darker background, is nothing less amazing than the first one

It's just slightly different color, slightly different lighting. But this is also hyper realistic in my opinion

So yeah, go ahead and try this hyper realistic prompt on something else

Could be a person, an animal, nature, landscape, anything you want really an object

Sky is the limit. Your imagination is really the limit. My next prompt is going to help you generate cute characters that you can use, for example

in a children's bedtime story or anything else that you like really

And the prompt goes like this. Create a cute broccoli character for my children's book

That was a very simple prompt. And here are the four broccoli characters that Dal E came up with

And they are very cute indeed. Look at this, like big eyes, shiny eyes

And they have slightly different artistic styles. So this one is more like Japanese anime kind of manga inspired

This one is more like I would say a classic children sketch

Not so much Japanese influence right here. Here the Japanese influence is coming back

You can see that mostly in the eyes and the way the eyes are round with the little light effects in them

And this one is also more like a watercolor kind of illustration

So four very distinguishable artistic styles here. All right. So that was the generate characters

cute characters prompt and I have tried this character prompt on some other things like for

example a fried tofu cube like this one right here an illustration of a playful and cartoonish

fried tofu cube with big expressive eyes and a cheerful smile looking ready for an adventure I mean

look at how cute this fried tofu is I mean who wouldn't bite it so there you go some more

fried tofu and different styles and this one reminds me of the little WhatsApp cat I don't remember

the name of that cat but the kind of like the little lines characteristic of this drawing style

and this one right here all kind of like very much cartoonish anime style so yeah there you go go have fun

with this cute character prompt this next one is about creating sets of WhatsApp stickers and it goes like

this create a set of WhatsApp stickers showcasing a koala character in different situations and expressing

different emotions so there you go I have my two different sets of stickers and

and with a consistent koala character. This is the first one. This is the second one

And I like that in the prompt, the koala is showcased in various situations

like sleeping on a tree, munching on eucalyptus leaves, playing with a toy

The stickers also capture different emotions, like happiness, surprise and sadness. I like that

ChatGPT added that dimension of the emotions in my prompt because really that's what stickers are for

when used on an app like WhatsApp. Take a look at the second one

We have different emotions or situations, scenarios like reading a book, dancing and daydreaming

like this one right here. All right, so generating sets of stickers also works very well

Now, of course, you can use this in a professional setting. Let's say, for example, I need some illustration for a blog post I'm writing

Well, you can prompt that E with the help of chat GPT to generate these illustrations for your blog post

a prompt like this one, for example, generate a few alternatives of blog post images to illustrate

a blog post about the future of remote working. That is the topic that you can change

I need them in a landscape aspect ratio 1920 by 1080 pixels So here in this particular prompt you can see two different variables the first one is obviously the topic of my blog post so that will depend on whatever you writing about but the second one is the aspect ratio and as far as i can understand how aspect ratios work right now in daly three and chat gpt you can decide between three different aspect ratio you cannot yet give them like a very precise aspect

ratio say I want a 5 by 4 aspect ratio I want like a 1320 by 1,800 pixel kind of visual

like you would do in Canva this is probably not going to work however you can still choose between

three main types of aspect ratio the vertical or portrait mode the landscape mode which is what I

ask here or the one by one or squared aspect ratio mode so that is the squared photo for

Instagram for example so portrait landscape and squared okay so stick to that and you should be fine probably in the near future we'll have

more precise aspect ratio dimensions resolutions coming we'll see that coming to dal e but as of today

stick with these three types of aspect ratio so here i chose the landscape and let me show you the

result for my blog post about the future of work so i have these two different images all right

so the future of work we can see two women here working in what seems to be a futuristic city

or office space without any glass or very transparent glass, a bunch of drones flying around and

screens, kind of huge screens or holograms of screens. We can see the city expanding very far

And yeah, this looks very futuristic. If you want to read the entire prompt, if you want to copy it

if you want to tweak very easily, you can do so from here. I can download my image

Now let's have a look at the second one. Quite different. We see an office space with a very nice representation of the famous

IKEA shelf like drawer shelf right here wooden desk what seems to be like an iMac screen

a MacBook Pro computer maybe the new Apple Vision headsets and a robot right here which is sitting

on the floor hardwood floor very nice chair very similar to my chair this is dreaming guys are you

looking at me chaty pt what the heck yeah so very cool block post illustration in the

landscape format i really like that and And yeah, I think this is well done

Not many hallucinations in here, not many big design mistakes. Maybe the leg of the pair of glasses here

it's not completely perfect. For the rest, I mean, this MacBook looks pretty cool

And this iMac screen also looks pretty cool. The keyboard, the Apple mouse

The future of work is clearly inspired by Apple. Just be aware of that

All right, let's move on to the next professional prompt. This time, let's say I need a visual

to illustrate a social media post. Let's admit we just reached 500,000 subscribers on our channel and we want to celebrate that

So here is a prompt you could use to generate that image. It goes like this. Generate a few one-by-one aspect ratio images to illustrate a post celebrating 500,000 followers on Instagram

I will post that image together with a post on my Instagram account

My audience is made out of fishing aficionados. All right. So I gave Dal E a bunch of different information

here and everything is between the squared brackets so you can change that information

I told it that I wanted a one-by-one or squared aspect ratio

I mentioned that the post is about celebrating 500,000 followers on which platform on Instagram

And I also gave the AI some guidelines about who my target audience, who my followers are

They are fishing aficionado. So now I expect the visual to be not only on brand

and on platform, but also to resonate with my target audience, with my follower base

So let's see what that he came up with. So these were the first two and you'll see that I reworked

this prompt. The first one right here says 5,000k followers, but it has a very nice

illustration going on here, like a lot of fishing gear, fishing rods, fishes and baits. So this is

really cool. I like the color scheme. The only problem is the text, right? It shouldn't say 5,000

but 500k followers. So that is a problem. And then the next one, which says 500k this time

so this is correct. And this is really nicely done with like fishing nets with the very nice

illustration here as well. So this one is correct. So I could use this one directly on my

Instagram account if I wanted to. But I wanted to try and fix these typos like these numbers

mistake on these images. So what I did is just click on the regenerate button right here. You can

always regenerate if you're not satisfied about the outcome and that's what I did and here is the

outcome I got the second time so these two new images that look great but also had some typos

some spelling mistake and again with the number followers seems to be understood by Dali it can write followers

but it cannot write numbers very well so here we see like 500 000 yes but the coma are a little bit funky

It should be only one comma. This one right here. This one shouldn't be there. So that's one

And then the other one, again, like problem with the number, problem with the coma

The rest looks great. Like, what a pity. And to be completely transparent, I tried and tried and tried to fix this

But it was very hard to get the right text to be displayed on my post

So I kind of gave up. I still had one out of five. This one right here that was correctly written

So I guess I could just fly with this one. All right. And my next prompt is about generating a logo for your business, for your brand or for whatever

you need. And here I wanted to give Dali a second opportunity to amaze me to blow my mind using text, right

So I wanted to go to the bottom of this. Can this thing generate logos or images that include text

Because when you look at Dali or Open AIs documentation or landing pages about this new technology

They showcase some beautiful posters and images that include text and there is no typo in their text

at least in their demonstration. So I thought it must be a way, so let's dig a bit deeper

Can I work with text in Dal-E? So that's the big question

So here is the next prompt I came up with. Give me a few options for a wordmark logo for my company called Fushis

It should be playful and modern. So that is the prompt. These are the two first attempt from Dalai to generate my brand logo, my wordmark logo for my brand

called fushis. All right. So we have the first one called fushes like the eye is missing

Fushis, I completely spelled the word fushis, but the eye is not here. The eye is missing on this one

And then we have this one, Fushis with two eyes. So obviously there's a problem here

So what I try to do and I recommend you do the same. Whenever Dal-E generates some images that are not perfect, you can try to steer back

the conversation in the right direction. So I would definitely continue the conversation here

and I said, okay, these look great, but the correct spelling is F-U-S-H-I-E-S. Could you adjust the spelling

in the previous design? One thing we need to understand is that Dal-E, at least as of today

is not capable of taking the same exact design and make one slight modification to it

It will regenerate another prompt, a similar prompt, but another prompt, and therefore the end result will be slightly different

All right So from these two images right here my original images I got these two right here You see they are slightly different This one is completely different But this one is also different

If we look closely, this was kind of like a round background with the blue kind of sky-inspired background

Now it's squared. The rest is pretty similar, but it's not exactly the same

You see the H for example, color is very different. So here I have Foohees now, not Fushis, Foohees

And here I have Foo-Shee's. she's so this one is actually correctly spelled but the e is a little bit funky looks like an

f or it could be considered as an e but it's a very different logo than this one so what i wanted to

show you here is when you ask dal e to reiterate to correct or to fix a specific image it will not

regenerate the exact same image it will create a new prompt and it will therefore generate a

slightly different version of your image but if you prompt enough times you might get something very

similar or you might get something even better so I was not completely happy with

this and I really wanted my word to be correctly spelled so I prompted again and

still got one out of two only and this one is not my favorite one I would have

loved this one to be correctly spelled and I'm sure if I continued prompting

enough times I eventually would have gotten it right I could have followed by a

prompt saying okay I really like the first image could you generate something

similar but make sure you write the name of my brand correctly and then spell it again

But at this time I was a little bit annoyed and I just gave up on trying to get it right

because I already had a couple right like this one is correctly spelled. So my conclusion about working with text, we are and we will probably get there

but as of today it's still a bit wonky and you will need a lot of patience if you want to get it right

All right guys, how are we doing? Are you enjoying this tutorial? If you do

I would appreciate if you give me a good old like so that we can push this video to more people right here on YouTube

We have a bunch of other videos about working with AI, chat GPT, mid-journey, etc

In our generative AI playlist on the channel, I will link that in the description for you

If you want to dig deeper, also I will have a card at the end of this video

But for now, I would like to wrap up the tutorial with a series of rapid fire tips on how to get the most out of Dali 3

And this is going to be really fast. Don't worry. The first tip is that the more details you provide in your conversation in your prompt

the better outcome you would get with DAL-E, all right? And we can see that on the DALI landing page right here

If we look at some of the examples that Open AI provides here, we have some very detailed

prompts like this one right here. If you click on the photo, you will access the entire prompt right here

And if you read it, you will see that this is a quite elaborated prompt

But the outcome is quite impressive. like this tiger divided between the fire and the green part, it's quite nice

Another one, for example, is this poster that, again, uses the exact good text and looks great

Maybe apart from what's here at the bottom, but Explore Venus is correctly spelled

And again, this is a quite detailed prompt. Let me give you one last example

This one I really love. It's really, really original. A landscape only made out of meat

Now, this is not for vegans for sure, but I thought this was a very original picture

that has been generated. Again, quite an elaborated prompt right here. So the conclusion of this is that the more details you give Dal-E, the better the image will be

Tip number two, and I already talked about this, do not hesitate to follow up, to correct your

prompt, to readjust the direction of where this is going, or simply use the regenerate button

to create some new variations of your image. Tip number three, ask more of the same

If among the outputs that have been generated by Dal E, there is one that you particularly like

ask the AI to generate more of the same. So, for example, say, I really like picture number one here

Can you generate more of the same? And it will do so. Tip number four, explain which aspect ratio you would like the picture to be generated in

Again, three different aspect ratio, squared, landscape, portrait. That's all we have for now

but let's hope OpenAI adds more and more precise aspect ratios in the future

Tip number five, you can zoom in or out on your picture

That is to have a different type of framing. So you could use keywords like close up

You could use keyword like mid shot, long shot, open shot, wide shot

What I would recommend is that you do a little bit of research about photography and the different types of framing and then use these keywords in your prompt to achieve specific results

And to make the lives of our members easier, I will add these keywords in the little canbat document I have prepared for you

And then finally, tip number six, you can ask that to modify your photo

You can ask it to change the colors, for example, to change the artistic style of a specific image

to change the medium. Maybe instead of watercolor, you want a hyper-realistic photo, or you want a vectorized graphic

whatever, that's the medium. Or you can also change the lighting style, like give some instruction about lighting to the AI

So changing any of these variables will result. in a different image

All right, so quickly before I leave you, I'm going to answer two questions that I can see coming

already from here. Number one, can I use these images for commercial use or for client projects or can I just

sell these images if I generate them? Well, the good news is that if you generate them with DALE3 in chat GPT, the answer is yes

And we can find this information right here, black on white, or let's say white on blue, on

OpenAI's website. It says as with DAL-E2, the images you create with DAL-E3 are yours to use and you don't

need our permission to reprint, sell or merchandise them. So this is very clear, guys, you can use

these images for commercial purposes. And then the last thing I want to touch on before I leave you

is can I really generate anything with DAL-E? The answer is no, because Open AI has put some

safeguards. There are some stuff that you will not be able to generate, because you will not be able to generate

they want to make this tool safe for anyone to use and not offend anyone

And again, you can read more about that on OpenAI's Dal E3 landing page

But basically there are three things you will not be able to do with Dal E

is to generate potentially harmful imagery, including violent, adult or hateful content

You will not be able to generate content in the style of living artists, images, or public figures

So you can forget about generate this in the style of of Banksy or generate this in the style of Picasso, this prompt will not be accepted

That is a smart way to prevent all of the copyright infringement issues because these artists

are not always happy that we can just emulate their style with a single prompt

So this here, this limitation of not being able to copy the style of living artists, images

or public figures is going to help prevent that, all right? And third, and lastly, you cannot upload images to be used as reference as of today

So you cannot upload an image of yourself and ask that E to turn you into a cute broccoli character

That will not work. There are ways around that. That might be for another tutorial

But for now, this is not something you can do. All right, guys, that is it for today

As I promised, for those of you who are members of our channel, meaning who are part of our paid membership

you will have access to a document. And in that document, there will be a bonus for you with 50 different keywords

And these are 50 different artistic styles that you can try while prompting that e

All right. So that's it for today. I'm going to leave you guys with our Generative AI playlist right here

in which you will find plenty of tutorials like this one. Thank you for watching. I will see you in the next video