How to Create Images with ChatGPT (With Prompts!)

Visual aids make things much more fun to read. But if you’re like me, all your photos are crooked because your sister’s dog pushed your tripod off a waterfall.

…no, just me?

Maybe you have another reason for not having stock photo quality images.

Whether you’re writing a blog post, sharing something on Facebook, or creating slides for a presentation, AI can help you make your content more engaging for your users.

I’m going to walk you through how to create AI images with ChatGPT, including some prompts and custom GPTs for you to try out for even better results.

We’ll also discuss the legalities of using AI images – although I’m not a lawyer – to help you understand the risks of using AI in some cases.

How to Create Images With ChatGPT

Step 1: Create a GPT 4 Account

You will need a ChatGPT paid account to access Dalle-3 which is currently only available in the GPT 4 plan.

When you’re using the paid account, be sure to have GPT 4 selected from the drop down, pictured above.

Step 2: Ask ChatGPT to Create an Image

To get ChatGPT to make an image, all you have to do is ask!

“Hey ChatGPT, please make an image of a golden retriever eating pizza.”

Give it a minute to generate the image. You’ll see the “Creating Image” loading icon show as it’s processing.

The more complex your image, and the busier the system, the longer it will take. It usually takes 1-2 minutes for me.

It’s that simple!

Step 3: Get Specific to Customize the Image

However to get the image you truly want, you’ll need to play around with your prompts.

For example, above I asked for a golden retriever eating pizza and got this photorealistic image:

But what if I want it to be a cartoon? I can either tell ChatGPT to make it a cartoon, or I can edit my initial prompt to be more specific.

If I tell it I want a cartoon of a golden retriever eating pizza, here’s what it gave me:

And what if I want it to add text on top of it saying something like “Pizza Yum!”. I would just tell it to add the text “Pizza Yum!” onto the image. Here’s the output:

There we have it!

Now ChatGPT can only generate one image at a time.

So if you want variations or options, you’ll need to reprompt it or ask for those variations.

But be careful! You’ll lose the image it already gave out if you edit the prompt above it. So always download the image if you think you may want to use it.

Step 4: Check the Prompt

Once you land on an image you love, you can actually see the prompt you should have given at the start to create it!

Click on the image and it’ll open in full view. Then click on the information button (an “i” in a circle) and it’ll open up the prompt to explain the image.

For the cartoon image we generated before, this is the prompt it gave:

“A cartoon-style golden retriever sitting at a table, happily eating a slice of pizza with a big smile on its face. The dog is depicted in a whimsical, exaggerated cartoon fashion, with large, expressive eyes and a wagging tail, emphasizing its joy. The pizza slice, loaded with cheese and pepperoni, is held in its paws. Above the dog, in a colorful, comic-style speech bubble, it says “Pizza Yum!” The background is a stylized, colorful kitchen, enhancing the playful and cheerful vibe of the scene.”

Did you notice how this is basically alt text?? And you can use it as that! I’ve copied these many times to use as alt text, although I’ll usually cut them down a bit (for example, the kitchen is irrelevant here so I’d remove those details).

You can use this is a base prompt to try to create more versions of the image or to replicate something similar in the future.

Step 5: Edit

Did you know you can edit AI images with ChatGPT?

This is a newer feature that I am so excited about.

Click on the image again to make it full screen. Then click on the little paintbrush icon with a circle around it. This is the “select” tool to edit the image.

Click and drag the select tool over the part of the image that you would like to edit. Let’s say we want to give our golden retriever green eyes.

I dragged the selector over the eyes, then typed in “make the eyes green”.

And here is our output, an image with green eyes:

Step 6: Download the Image

Once you’re happy with your image, you can save it.

Click the download button (an arrow pointing down to a flat line) on either the preview of the image in the chat like this:

Or in the full screen preview of the image like this:

The image will save to your computer as “Dall-e [Date] [First sentence of prompt for the piece].webp” in your Downloads folder.

Step 7: Resize the Photo

The images are 1,024px on the shorter side (1792px on the longest side if it’s a rectangle).

Now, for Google Discover and for most large scale presentations, that’s too small. We want at least 1,200px for Google, and even larger for a giant slide deck.

I take the photo over to Canva to edit it. I will create a canvas that is whatever size I want, let’s say 1,200px by 1,200px for our square pizza dog.

Then I drag and drop in the image, and it automatically resizes.

Now I can download the image as a jpeg for my website, or a png for a slide deck.

I can also rename it so it’s something more useful like “cartoon golden retriever eating pizza yum”, so it’s not an insanely long name.

This also technically helps to wipe the metadata of the AI image, but that’s not my main reason for doing it. I just want a bigger image, and this has proven to be the best way to reduce grain when resizing these images.

(Some people speculate it’s good to do this as Google may not like AI images, but we have no concrete proof of that yet from Google or any explicit statements on it.)

You can also further edit the image in Canva by adding more text, putting in a frame, or otherwise making adjustments until it’s what you need.

Now you have a beautiful image to use for your content!

Video Demo

Note: This was a previous version of Dalle-3 that used to give 4 outputs. Now you can only get one output at a time.

Prompts to Create ChatGPT Images

To create a great prompt for ChatGPT images you need to know a few things:

1. Style (think photorealistic, watercolour, etc.)
2. Aspect ratio (even just horizontal versus vertical)
3. Number of subjects
4. Point of view of the image
5. Where the subject appears in the image
6. Background
7. Tone or emotional atmosphere
8. Colours to include
9. Specific lighting instructions
10. Text
11. Any examples or references

The more specific, the better your image will be, and the more custom it will be for you.

Here are some example prompts to help you understand:

1. Create a photo realistic stock photo of a blonde woman from Finland working remotely at a desk with a silver laptop. The image should be portrait with a single subject, a white office in the background, and the person should be wearing pink headphones. The subject is facing the image front on.
2. Create a hyper detailed oil painting of a golden retriever dressed as a knight in shining armour leading his army into a battle with another kingdom. He carries a long silver sword and has a paw print on his armor over his heart as he calls out to his army. There is rain pouring down and everything is dark and moody.
3. Create a photo realistic, panoramic image that shows the essence of New Zealand travel. Include people hiking up a mountain at sunset. The sky is tinged with pink and reds, reflecting over the distant ocean. The image should be 1792×1024 pixels.
4. Create a beautiful landscape featuring a calm lake on a clear blue day, with mjestic mountains in the background. The scenery should be reflected on the water’s surface.
5. [Paste a blog post]. Please create a featured image based on this blog post. It should be horizontal with a simple background.
6. Create two side-by-side panels in an animation style. The first panel shows a man looking haggard at his work desk really depressed. The second panel shows the same person but exuberant and jumping up. Use vibrant textures in the background to make the man pop and convey that he’s excited.
7. Create a realistic cake with a magical forest scene complete with miniature fairies and fireflies, and ethereal woodland creatures.
8. Create a series of affirmation cards about overcoming anxiety. They should portray women in real scenarios, specifically around travel. Keep the illustrations on each card minimal with a plane background. It should be a flat illustration with a white background and vertical images.
9. Create a cubist image of a grizzly bear catching a salmon during spawning season in a river. There should be water flying and the image should be centered on the grizzly bear. The image shows the untamed nature of the wild and the beauty of the natural life cycle.
10. Create a chibi-style illustration of a shy woman with her hair braided wearing a flowery dress with red flowers and brown sandals. She is smiling shyly with red blush spots on her cheek in front of a garden of roses. The camera angle should be low to show off her shy posture.
11. Here is a picture of my dog. [upload picture]. Please make him a Peanuts’ style cartoon.

Custom GPTs for AI Images

If you don’t want to worry about prompting ChatGPT too much, some lovely creators have done the work for you!

You can find custom GPTs in the GPT store (GPT 4 subscription required).

Here are some I’ve enjoyed:

1. Food Photography – To make realistic images of food for food bloggers. Works great for drinks too!
2. Pixar My Pet – Create a Pixar style movie poster of your pet for fun.
3. Photo Realistic GPT – Create photo realistic image of a person.
4. Logo Creator – Generate a vector style logo for your business.
5. Cartoonize Yourself – Turn yourself into a cartoon (think Up!).
6. Super Describe – Upload an image and have ChatGPT give you a prompt to recreate it or create something based on it.
7. Drawn to Style – Transform a sketch into another style. Take your doodles and turn them into masterpieces!
8. Custom Character GPT – Generate custom character that you can reuse and repose with this GPT.

Limitations of Creating AI Images

AI images are not perfect. No AI generator is able to be as precise or customizable as an actual illustrator. So keep that in mind when you’re working with them!

AI image generators are notorious for being extremely flawed with text. In fact, I’m surprised my “Pizza Yum!” text turned out properly on the first try. Usually it takes many more tries, or even adding the text in manually in Canva.

The same goes for numbers or icons on an image. The more you add, and the more complex an image gets, the more likely you are to have an error.

You can only generate one image at a time. Initially, it would show you four images and ask you to select the one you liked best. This was to help train the AI on what a good output is for each prompt.

Now that it knows better, they only provide one image per prompt.

ChatGPT also limits the number of images you can make an hour. When they first launched Dall-e 3, it was about 20 per hour. Now you can get more like 50 an hour before it hits its cap. This will likely grow with future updates. But keep it in mind if you’re messing about with some prompts but have a real deadline you need to meet within the hour.

I find if you customize one element, especially outside of the edit function, ChatGPT will undo other edits or even change the whole image entirely. So be careful to save any image before you make adjustments or undo a prompt.

ChatGPT doesn’t like to create things in the style or likeness of copyrighted works or famous people. So if you asked it to make Robert Downey Jr. in a blue wig, it will usually protest. You can usually encourage it to do it by telling it it’s just for fun, and not for redistribution.

But, that brings us to the main issue with AI images…

Legalities of Using AI Generated Images

AI images cannot be copyrighted.

Just like AI generated writing can’t be copyrighted by you if you didn’t edit it, AI images are created by a non human entity. The US Supreme Court has made rulings in the past, most notably about a monkey taking a selfie in Bali, that non human entities cannot hold copyright, being non human. And a human cannot claim the copyright on their behalf.

So you telling ChatGPT to make a pizza eating golden retriever, doesn’t mean you “created” it. You didn’t draw it, so you can’t own it.

There are also legal concerns over how ChatGPT and all AI image generators train their models on art from artists who did not consent.

If the AI generates art that is in a particular artist’s style, there could be legal issues if someone ever manages to hold an AI image generator accountable for their training models.

Personally, I don’t feel comfortable selling AI art for that reason. Some people use it to create stickers or planners to sell, but I don’t.

I’d much rather use it for fun – like the odd featured image for blog posts or for some cartoon images of my service dog in my slide decks for presentations that would otherwise just be bullet point lists.

Those in and of themselves are not making me money, and they’re easy to remove if there ever was a litigation issue.

I’ve also used the images on Facebook in my free FB group to have people like the image in order for me to “make the blogger more successful” (she very quickly ended up in outer space then became sentient gas? It was wild!).

That was just for fun.

I also love turning my dog into a Pixar movie poster for my own purposes – like this:

View this post on Instagram

A post shared by Nina Clapperton (She Knows SEO) (@ninaclapperton)

In my opinion, AI is best used for entertainment purposes, not for monetization at this point.

Until we have clearer legislation and protections for artists, I wouldn’t risk selling these images and I don’t recommend you risk it either.

That being said, you’re a grown up so do what you want. I’m not your lawyer – or anyone’s lawyer. Thank god for ditching law school for a $200 flight to New Zealand!

Conclusion

Now you know everything you need to start generating your first AI images.

It quickly becomes addictive turning all of your friends and family into movie posters, or creating illustrations of your favourite travel photos to hang on your wall.

I’ll be the fun police one more time: be careful with AI images because we don’t know the long term legal ramifications of using them yet.

Don’t base your whole business on AI images and you should be ok (fingers crossed!).

Mostly, have fun using them and experimenting with prompts. I’ve seen 1 page prompts in Facebook groups that generate images so incredible that I want to go and live in that world!

I find studying the outputs ChatGPT gives me for the images it creates is the best way to learn the structure and cadence to a proper AI image prompt.

It’ll also help you get better at spotting bad AI images – there are so many out there, especially on Facebook.

Now that you know how to make AI images, learn how to write newsletters with AI! You can include AI images in your newsletter for fun engagement with your audience.

Read More About AI and Content Creation

• 28 Best AI Copywriting Tools (SEO + Blogger Tested!)
• Which AI Writer is Actually Worth Your Money: ChatGPT vs. Jasper.ai
• What is SEO Writing?: Beginners Guide to SEO for Blogs (with Examples!)
• How to Make Money as a Travel Blogger: 15 Ways I Made $30k
• Honest RankIQ Review: Is It Worth It for SEO?
• 19 Best Surfer SEO Alternatives for Bloggers
• Link Whisper Review: SEO Expert’s Take on the Internal Linking Tool
• How to Automate Repurposing Your Blog with ChatGPT and Zapier
• 10 Surprising Uses for ChatGPT You Haven’t Tried Yet
• How to Define a Style with ChatGPT to Write a Newsletter
• How to Create Images with ChatGPT (With Prompts!)
• How to Make Custom GPTs in 4 Steps
• How to Use NotebookLM to Make a Podcast From Your Blog Post

Like this post? Share it with others!