How To Use AI Images for Creative Inspiration

Images generated by artificial intelligence (AI) can be a great way to express creative ideas. With the help of an AI image generator, you’re able to visualize your thoughts and inspirations in nearly any format. Want to see a painting of a UFO in the style of Leonardo Da Vinci? No problem. AI engines can make that happen.

When you use an AI tool to create art, you can:

• Translate your written ideas into a visual medium
• Create unique backgrounds and elements to layer into your original digital art
• Take art you’ve already created and make digital tweaks to it, uncovering new versions and variations
• Simulate what different photographic styles look like based on year, camera, and film type

You don’t have to be a computer expert—or an artist—to take advantage of using AI images for creative inspiration. In this guide, you’ll learn about three of the most popular and robust AI image generators available today. We’ll also cover everything you need to know about generating AI images, including:

• What are AI image generators?
• How do I start creating my own AI art?
• How do I use AI images?
• Best practices for using AI images

What are AI image generators?

AI image generators tap into artificial neural networks—large, data-rich systems that act as the “brain” of the AI tool. These systems are trained to recognize different elements of images and understand how they relate to words.

Because these artificial neural networks are not yet trained on every single possible word and image pairing, there are times that you’ll wind up with results you weren’t expecting.

However, by spending time developing the right prompt, you can help the AI get reasonably close to what you want. Try it yourself by following along as we explore how Midjourney, DALL·E 2, and Stable Diffusion work.

Midjourney

Midjourney is an independent research lab that’s developed an AI-powered text-to-image generator. The tool integrates with Discord, a chat app.

Everyone using Midjourney submits their prompts in an ongoing Discord conversation thread, meaning you can see what other people are generating—and they can see your images in turn.

This makes Midjourney uniquely interesting from a creative standpoint. If you’re looking for creative ideas, all you have to do is log into Discord and watch the rapid stream of images fly by.

Midjourney offers a free trial, and there’s no charge to use Discord. Anyone can log in and begin making images.

While it is possible to ask Midjourney to modify an existing photo, the chat-based interface makes it a little more complicated than do other AI image generators.

If you’d like to generate entirely new images from text prompts, though, Midjourney is an excellent option. It’s easy to input text, review the results, and ask Midjourney to make tweaks.

DALL·E 2

At the time of writing, DALL·E 2 is the latest version of OpenAI’s image generator. Because it’s created by the same company that produced ChatGPT, you might expect that DALL·E 2 has a good grasp on outputting text—but it doesn’t. If you try, you’ll get a lot of typos.

‍

The DALL·E 2 neural network is best at reading what you input and producing an image. It’s particularly good at replicating different art and photography styles.

Anyone can use DALL·E 2 through their web browser, and all users get a monthly bundle of free credits to experiment with the AI.

Stable Diffusion

Stable Diffusion was created by Stability AI. It’s very geared toward digital artists. You can tap into Stable Diffusion directly through Adobe Photoshop and Blender (a 3D rendering tool) with the help of plugins.

However, the use of this tool isn’t limited to digital artists. Anyone can use Stable Diffusion through a web browser. While several third-party interfaces use an API to access Stable Diffusion’s neural network, you can access the official beta through DreamStudio.

Stable Diffusion is a good choice for anyone who wants to have a lot of control over how they use AI to generate works of art. DreamStudio lets you tweak many aspects of the process, making it very simple to either upload a starting image or recall one of the AI’s prior outputs.

How do I start creating my own AI art?

You can start creating your own AI art with any of the tools in this guide. You can take advantage of each tool’s free trial to determine which fits your needs best.

To get started, jot down some ideas you’d like to see in image form. These could be story ideas in your head, scenarios a client has described (that you want some inspiration for), or a scene you read in a book.

You can also browse different fine art movements and styles to get inspiration—or visit a local art museum! AI art generators can recognize and replicate many famous art movements and artists’ styles.

When you’re ready to begin generating, pick one of the tools in this guide and follow our step-by-step prompts to maximize your free trial time and determine which one you’d like to keep using.

One quick note—AI is constantly evolving, and many AI tools are still in beta. While all of the instructions in this guide are accurate at the time of writing, you may find that your experience is a bit different. You can always find up-to-date information about each tool here:

• Midjourney Quick Start Guide
• DALL·E 2 Features and Updates
• Stable Diffusion Prompt Guide

Unleashing your creativity with Midjourney

Midjourney is an excellent tool for generating images that closely mimic other art styles or depict fantastical scenes. It’s a bit more complicated to set up than the other AI image generators on this list, but once you know how to use Midjourney you can create some fantastic images.

Step 1: Sign up for Midjourney with a Discord account

To get started, visit the Midjourney website and click “join the beta.” A new window will open, asking you to create or log in with an existing Discord account.

Discord is a free chat app similar in many ways to Slack and WhatsApp. You can have one-on-one conversations as well as group chats. Midjourney uses the latter.

Once you’re signed in, you’ll have Midjourney and Discord accounts. You can access Midjourney through Discord’s web or desktop application.

‍

Step 2: Enter a prompt into the Midjourney Discord chat

To use Midjourney, you have to enter your prompt as a chat message in Discord. These messages—and the resulting images—are public by default, and other people can download or make variations on the images you generate.

Start prompt by typing /imagine into the chat box. Then add a sentence or two describing the image you’d like Midjourney to create.

I asked Midjourney to create an image of “a cat working from home”. It gave me four similar pictures in various styles:

‍

I found Midjourney’s output to be very accurate right away, with minimal adjustments needed on my end.

Step 3: Experiment with different art styles

Midjourney is capable of replicating specific art styles with impressive accuracy.

I like paintings done by the Dutch Golden Age masters—and when I asked for a cat working from home in this style, Midjourney delivered:

‍

I also took inspiration from other image prompts I saw in the Discord chat. Another user was asking for “polygons” style images that I thought were very interesting—so I tried it too:

‍

And, by asking for a sticker of a cat working from home, I was able to generate these fun variations:

‍

Step 4: Make tweaks to images with Midjourney AI

If Midjourney gives you a picture, and you like parts of it but not the whole thing, you can re-do your request and add a negative prompt onto the end.

To do this, type two dashes followed by “no” at the end of your prompt. Then, type out what you’d like to remove.

Midjourney also lets you continue to iterate on one of the outputs it produces. When you get your first set of four images from Midjourney, you’ll see a series of controls at the bottom of the image.

‍

The image outputs are numbered clockwise in quadrants, so if you click “V1” you’ll get three additional variants of the image in quadrant one.

‍

If you click “U1” you’ll get a larger version of that image. This is called “upscaling”.

‍

The original image you base these changes on is called the “seed” image.

You can also make variations of, or upscale, other people’s images. It’s important to note that if you do so, you don’t own the resulting image per the Midjourney terms and conditions.

In fact, you only own the images you create if you’re a paying Midjourney member … and Midjourney can reuse the prompts (including images) that you feed into its AI.

However, all users, including ones on a free trial, can use Midjourney images under a Creative Commons license.

Step 5: Organize, download, and reuse your images

You can go to Midjourney’s website and see all of your images at any time—they get logged in your account dashboard. From there, you can re-use an image, turn it into a seed, and organize your files into collections.

‍

Sample Midjourney prompts to try

Midjourney is quite good at understanding text prompts, and you can generally give it instructions in any order. Some users give Midjourney long paragraphs of text, but the AI can’t always keep up with that level of detail. It’s best to use short, concise prompts that are one to two sentences long.

A good Midjourney prompt format to begin with is:

/imagine + subject of image + action + environment + art style + other descriptors - no (elements you want to remove)

Give it a try by typing /imagine and then one of these example prompts:

• A solar eclipse, and the moon has rings around it
• A dog astronaut floating in cyberpunk space, light pastels - no stars
• A hyper-realistic fairy in a futuristic urban woodland scene
• A dog working from home, done in Kawaii illustration style

If you’d like to use an existing photo as a seed for your Midjourney generation, you can upload the photo into the Discord chat, copy a link to the photo message, and include it right before your prompt.

Capturing photo-realism with DALL·E 2

If you’d like to replicate specific camera styles or generate photorealistic results, give DALL·E 2 a try.

Step 1: Log into the DALL·E 2 interface

To log into the DALL·E 2 interface, you’ll need an OpenAI account—if you already have one you use for ChatGPT, that’ll work here as well.

Step 2: Enter a text prompt

You’ll need to be very specific about the type of output you want—if you just say “image,” DALL·E 2 sometimes disregards later clarifications about style.

For example, I typed “image of a cat working at home, photorealism” and DALL·E 2 gave me hand-drawn style images, not photos.

‍

When I modified my input to say “photo of a cat working at home,” I got a more photographic output.

‍

Step 3: Modify the output with additional parameters

This is where DALL·E 2 gets really fun. Once you’ve settled on a text prompt that’s returning the general style of image you want, you can further modify how it looks by adding details about different camera and film types.

Here’s an example of DALL·E 2 returned when I typed “photo of a cat working at home, Nikon 35mm lens”:

‍

It’s not perfect—the keyboard and screen are warped—but it did capture more of that depth of field that I’d expect to see in a real photo.

I wanted to see if DALL·E 2 could replicate older cameras, so I asked it to give me an image in the style of an Olympus Pen-EE half-frame camera from the 1960s. It returned a few options, and one was pretty close to what you might get out of that camera!

‍

Step 4: Continue tweaking the image with AI

If you wind up with an image you like but it isn’t quite perfect, you can ask DALL·E 2 to keep tweaking that output.

All you need to do is click the three dots in the upper right corner of any image generated by DALL·E 2 and select “generate variations.” You’re giving DALL·E 2 a seed image, just like in MidJourney.

‍

The AI will then produce four new pictures based on this seed.

‍

DALL·E 2 doesn’t understand negative prompts very well—but you can have impressive control over how the AI edits your images.

Start by clicking on the image you’d like to modify, and then click “edit.” (You can also create variations from this screen!)

‍

In the window that opens up, use the eraser tool to remove any parts of the image you want to change. You’ll then need to give DALL·E 2 a new prompt for what the entire image should look like—not just the part you’re modifying.

‍

DALL·E 2 will then modify the area you erased and incorporate the changes into the seed image.

‍

Once you’re happy with the images produced by DALL·E 2, you can download the files to your computer for further editing or sharing. OpenAI says you can use content generated by its tools for any purpose as long as you agree to the rest of its terms and conditions.

Sample DALL·E 2 prompts to try

The formula for creating a basic DALL·E 2 prompt is simple:

Type of image you want + image subject + image action + image location + additional style modifiers

You can try creating prompts right now using variations on that formula, or copy and paste one of these examples into DALL·E 2:

• Photo of a cat eating dinner on the moon
• Oil painting of an astronaut tending to a garden
• A robot hiking on the moon in the style of Van Gogh
• Photo of two dogs playing in a field of daisies

‍

Fine-tune your art with Stable Diffusion

Stable Diffusion offers additional control over your AI generated and seed images, which makes it a nice choice for fine-tuning art. You can use it to:

• Add elements to art you created offline or with a non-AI software tool
• Further enhance AI art created by Stable Diffusion or another generator
• Modify existing source photographs

Step 1: Sign up for a DreamStudio account

Log into DreamStudio through the Stability AI website—you can sign up with an email, Discord, or Google account.

Step 2: Enter a prompt into DreamStudio

To start, you’ll want to type in a prompt, just like we did with Midjourney and DALL·E 2. Stable Diffusion handles long, highly specific prompts well—you can get inspiration from the Lexica Stable Diffusion prompt search engine.

I started by requesting an image of a cat working from home. It was fine, if a little basic:

‍

I poked around on Lexica for a bit and found an image I liked—a pretty room with lots of leafy plants. I then modified the Lexica example prompt to involve a cat working from home:

Architectural Digest photo of a cat working from home at a computer in a maximalist green solarpunk living room with lots of flowers and plants, golden light, hyperrealistic surrealism, epic, stunning

This increased level of detail resulted in an output that I was happy with … although the cat doesn’t appear to have any eyes. I could still use this image as inspiration for other work, though.

‍

Step 3: Fine-tune your art with Stable Diffusion AI controls

Once you have an image you like, you can continue modifying it with Stable Diffusion’s controls.

To start, click on the image you like. This will make the image preview larger, and log the file’s ID number as a new seed.  

‍

Use the sliders along the left to make adjustments to your image. You can change the:

• Image size
• Number of images generated
• The CFG scale, which indicates how closely the output should adhere to the prompt
• Steps used for each output, which can improve the quality of the image by reducing noise (if an image looks grainy, it’s very “noisy”)

Changing the size of your photo will give you a larger image with some retained elements from the seed.

Here’s another photo of a cat working from home that I generated with Stable Diffusion:

‍

Here’s what I got when I used that image as a seed and expanded the size of the output:

‍

As you can see, there are some similar elements—the presence of a yellow office chair and greenery on the walls.

You can use Stable Diffusion to make similar changes to an image you photographed or sourced elsewhere, too.

Take a look at the three images below. I snapped the left-most shot of Malarrif Lighthouse in Iceland. I thought the lighthouse looked a bit like an Atlas rocket, so I played around with AI generator prompts (using words like rocket, ship, space, planets, lunar, and landscape) and wound up with two different space-age scenes.

‍

You can still recognize some of the original elements of my photo, but it’s certainly not a lighthouse anymore!

Step 4: Further adjust your art with prompt weighting

Stable Diffusion also allows you to give weight to your prompts. This means you’ll tell the AI to prioritize certain aspects of your prompt over others.

To indicate that a portion of your prompt is more important, put a “1” after the relevant word. To indicate that part of your prompt can be deprioritized, put a “-1” after the relevant word.

You may need to play around with weighting different parts of your prompt to find just the right output. Check out the differences between these outputs based on how they were weighted:

‍

‍

The changes are subtle, but result in a better image. When I emphasized the cat and de-emphasized the chair, I wound up with a more realistic selection of cats in offices (not to mention computers that appear to be usable).

Sample Stable Diffusion prompts to try

For the best results when using Stable Diffusion, give the AI lots of detail in one or two sentences:

Specific example of a reference style + details about color and design + subject + action + location or environment + image quality modifiers

Put these example prompts into Stable Diffusion and see what you get!

• 90’s vintage anime illustration, surrealism, detailed line art, fine details, neon. A woman reading a book in a library.
• Disposable camera image, muted pastels. Family standing in the hall of mirrors, Versailles. Light flare.
• Hyperrealistic, highly detailed digital art. Woman in an astronaut suit on mars. Smooth soft light, high definition.
• Cubist painting, neon, glitter. A man playing a violin in a subway station.

How can you use AI images?

AI images aren’t perfect. But they’re good enough to help get the creative juices flowing—or a way to just have fun experimenting with what artificial intelligence can do.

Try using AI images during your creative process in one of these ways:

• Coming up with different variations on an idea
• Mocking up placeholder images for a website design
• Getting inspiration for what a photograph might look like if you turn it into a painting or a drawing
• Building storyboards for a film or illustration project
• Visually communicating your ideas for a project to other collaborators or stakeholders
• Creatively brainstorming in new ways

Why AI images aren’t a replacement for traditional illustration and art

At present, AI images aren’t a realistic substitute for traditional graphic design, illustration, or other forms of visual art.

There are a few reasons why. First, there’s the matter of copyright. AI image generators can’t actually think, create, or ideate like a human, even if it feels that way when you’re using them. These are computers trained on large data sets and existing content.

Sometimes, AI technology uses portions of photographs and illustrations created by working artists … without their permission. And that can create legal challenges. In January 2023, Getty Images announced a lawsuit against Stability AI, claiming that Stable Diffusion inappropriately accessed and used licensed Getty photos.

AI images also often contain errors that a human artist wouldn’t make. Take this Midjourney cat image, for example:

‍

While the image looks great at a distance, zooming in reveals that the cat’s glasses warp and disappear into its eyes. While this is a fine image to use for inspiration or as a placeholder, it’s not high quality enough to deliver to a paying client as finished work.

Lastly, when you produce an original piece of work you own it (or your client does, if they’ve paid you for the rights to the content). The ownership around AI-generated images is more nebulous, with AI image generation platforms using a mix of partial and Creative Commons licenses.

The U.S. government is currently exploring who gets the copyright for AI images. Until there are concrete answers, AI image ownership will remain a gray area.

‍

Best practices for using AI images

When you use AI-generated art in any project that you’re going to share with the public or a client, you’ll want to:

• Be transparent about the use of AI. If you’re going to use AI in a project (such as for creating placeholder images) let your collaborators and clients know that these are AI-generated reference images, not your original work.
• Give credit to AI image generators. Should you decide to include an AI image in any of your finished or published work, give credit to the tool you used—just as you would give credit to the original creator of an image you sourced through other means.
• Read the terms of service in full. Yes, reading terms of service agreements can be a drag, but this is an instance when you probably should read it and not just blindly check the “I agree” box. Each AI image generation tool has its own terms around how and when you can use the tool, as well as ownership of the outputs.
• Consider potential privacy implications. AI systems have to process and use the information you input in order to create an output. However, some of them may retain access to your inputs after you’re done—and have the ability to reuse them. Be clear on how your inputs will be used, and check with any clients and collaborators before feeding the AI their work or information.

By remaining mindful of the above—and staying abreast on any changes and developments to your favorite AI engine—you can develop a great workflow that uses AI tools to enhance your creativity.

Offer your services as an AI artist on Upwork

Once you’re comfortable using AI image generation tools, you can consider using Project Catalog^TM to offer your expertise to others. Examples of services you might provide as an AI artist include:

• Creating placeholder graphics for a web designer
• Helping writers begin exploring how to express their ideas visually
• Developing the images for a filmmaker’s storyboard
• Working with small business owners to create various logo concepts

To get started, log into your Upwork account to create a generative AI project and begin connecting with clients who need help. (You can also look to Talent Marketplace^TM  to find work as an artificial intelligence artist.)

And if you’re in need of someone who can help you harness the power of AI tools for your creative work, you can post a job and hire skilled AI professionals and AI designers right here on Upwork. Just log in with your Upwork account to get started.

Disclosure: Upwork is an OpenAI partner, giving OpenAI customers and other businesses direct access to trusted expert independent professionals experienced in working with OpenAI technologies.

Upwork does not control, operate, or sponsor the other tools or services discussed in this article, which are only provided as potential options. Each reader and company should take the time to adequately analyze and determine the tools or services that would best fit their specific needs and situation.