Learn how to harness the great powers of AI through this ChatGPT and Midjourney tutorial.
We will be building better prompts and making the most of these new technologies.
Let’s imagine Chat GPT 4 as a wise wizard trained for many years.
We can go to the wise Wizard and ask him to define us and articulate a specific prompt for different contexts.
To perform this process, we will train ChatGPT 4 in the art of prompt writing and help refine its outputs after this.
We’ll look at how these prompts come out in Midjourney and look for ways to improve our prompt writing wizard.
So let’s dive right into our ChatGPT and Midjourney tutorial.
And what we’re going to do is, we’re going to prime ChatGPT4 to act as our wizard prompt writer.
And to do that, I’m going to inform it that,’ You need to act as a prompt engineer.
You will write me prompts for an art generator called Midjourney.
I will provide you with short content ideas and your job is to elaborate them into full explicit coherent prompts.’
So, here I’ve defined the task.
I’ve told it what it’s going to act out and now I’m giving it more context around what an AI art generator is, what a prompt is, and how I would like the response formatted.
Prompts involve describing the content and style of images.
It is useful to be explicit and use references to popular culture artists and media.
Your focus needs to be on nouns and adjectives.
This prevents ChatGPT from going off and describing in flowery ways.
The vacant gaze that somebody’s having that reminds him of the first day of autumn.
And we’re asking Chat GPT to define what would that look like.
That the eyes are focused on something in the distance rather than a vacant stare.
I love to think of these prompts as casting spells into Midjourney.
With the right combination of words, we can elicit infinite possibilities.
What I’ve done in this ChatGPT and Midjourney tutorial so far is to break the prompt down into specific variables.
So that we can make sure we get adjectives relating to these different components of an effective prompt.
This will make sure that we have accurate and defined language for the whole ray of possibilities of our prompt.
When giving a prompt, remove the brackets, speak in natural language, and be more specific.
Use precise and articulate language.
Always output me to full prompt options.
This means it will get two options every time we ask for a prompt.
So this can be great for simultaneously creating different images inside Midjourney.
It also gives us a little bit more flexibility to either choose one or the other or even combine parts of both into our unique prompt.
Finally, I’ve given it one example prompt to give it an idea of how I would like the output to be to get the most out of ChatGPT.
It’s important to give it a lot of contextual information.
ChatGPT4 now takes up to 25 000 characters in contextual information.
This allows us to give it a lot more information which then it can include in its intelligence.
We can now ask it to reflect and interpret that in original ways.
An interesting use case that I recently saw was, somebody took a new feature from a coding language and inputted the entire documentation into Chat GPT.
Then he asked it to spit out a tutorial that was aimed at beginners.
So, we’re taking advanced complex language and we’re asking Chat GPT to output us a more accessible version of that information.
So, let’s take a look at the content that we’ve got from Chat GPT for our first prompts.
The first one is, a tranquil underwater landscape medium oil painting style inspired by JMW Turner and H.R Geiger.
Turner was an English painter who specifically worked in oil paints and created very beautiful atmospheric seascapes.
Now, the first thing I noticed is that it still included the brackets, so I will re-train ChatGPT by asking it to rewrite the prompts without brackets.
Okay, so this is looking much better.
What we’re going to do is we’re going to copy this prompt and we’re going to hop into Midjourney.
We’re going to make sure that we’re using version 5 for the algorithm and we’re going to type in forward slash imagine, and paste in our brand new prompt.
And this is going to spit us out something quite remarkable.
And simultaneously we can come back and do the second idea as well.
Using ChatGPT to Generate Personalized Prompts
At this point in our ChatGPT and Midjourney tutorial, we’re going to hop back into Midjourney
and see how it’s getting on with our original requests.
So the first prompt we put in was:
A tranquil underwater cityscape oil painting by JMW Turner with warm ethereal sunset lighting filtering through the water. The colors should be rich blues, teals, and warm oranges. Capture the scene with a wide-angle lens.
And it’s created a pretty interesting concept piece here.
Let’s upscale a couple of these.
Wow! It’s fantastic how it’s got this juxtaposition between the underwater scene and the cityscape on top.
And the second prompt I put in was a futuristic garden party with robots and people mingling in the style of Alfonso Mucha who is a very popular artist.
He had an art deco style and created several stylistic advertisements in the early 20th century.
Also please forgive me for the pronunciation of his name.
Futuristic Garden Party
Let’s get back to our Alphonse Mucha robotic garden party and upscale a couple of these for a closer inspection.
You can see that we’ve got this very distinctive interesting concept of a robotic party in the style of Alphonse Mucha.
I like it but I essentially try the prompt that we asked it for.
So, now we’re going to look at our vintage travel posters.
I’m also going to change the aspect ratio for this and make it vertical.
So for that, I’ll change it to dash dash AR 2×3 which means that we’re going to have a poster outputted.
For every 200 pixels of width, we’re going to have 300 pixels of height.
Now we’re going to go again with the second option and try that again with a 2×3 ratio.
In the meantime, we’re going to whack another one inside of ChatGPT.
New York Portraits
This time we’re going to go for street photography portraits in New York.
Midjourney version 5 has taken the levels and quality of AI art to new heights, specifically photorealism.
It has increased immensely the details of the finer elements of human anatomy.
They are now so precise that it’s becoming harder and harder to distinguish between AI art and real photography.
One thing that was rendered particularly well inside of Midjourney now is hair strands.
Look at these beautiful vintage-style posters that we’ve now gotten from Midjourney.
These would work really well for home decor.
Aside from this ChatGPT and Midjourney tutorial, let’s look into making money out of these stunning artworks.
Etsy Shop Idea
If you’re interested in a business idea, I can suggest creating very specific travel posters for different locations.
Then, open an Etsy Shop really targeted at a specific place that has a high number of rental properties.
Airbnb renters love to put specifically inspiring artworks that are relevant to a space on their walls.
I think that it is a great opportunity to monetize a piece of work like this.
Look at that one
I adore it.
The colors and the composition are fantastic and I really like the difference we’ve got between the two options.
The first is slightly more saturated with these brighter reds and the second has a more nuanced approach with these more pastel colors.
Letterings in Midjourney Needs Improvement
Now you may have noticed that the text is being rendered pretty horrendously inside of Midjourney.
This is something that hasn’t been improved very much in version 5.
One thing they did mention is that they are trying to remove unnecessary text where possible.
There is some notion of that happening, there is less text appearing in a lot of places.
But what we can always do is take an image like this and help ourselves stand out by actually adding our text.
You can use a free tool like Figma to easily add your text.
You can paste an image in easily, and what we’re going to do here is we’re just going to crop the image by drawing a frame and pasting the image inside this frame.
Then we’re going to extend the image upwards to give us some room at the top to input our text.
We’re going to put it in our writing.
I’m going to try Chamonix here.
Pick a suitable font.
I’m going to choose Poppins this time.
Make the font size up to 160.
Change the font style to bold, and we’re going to select a color from the inside of the artwork.
Make sure that we’ve got the whole word on one line, and I’ll slightly increase the letter spacing.
When you’re using all caps, it’s important to slightly increase the letter spacing.
You can quickly take an image and vastly improve the quality that you have by adding your text.
This is something that AI is massively struggling with still.
And it’s very interesting to see that it has overcome the issues with hands and guns yet the text is still something that it is suffering with.
Unable to Recreate Elements and Use Them in Future Works in Midjourney
Another area the V5 is still not able to deal with is being able to create specific elements, remember them, and put them into future works.
For example, you could create a character and then reuse the same character in the future.
The same goes with objects and locations, but that’s something hopefully we’ll see before too long.
But here is my vintage travel-style poster generated with Mid-journey and Chat TBT.
But let’s have a little look at how we’re getting on with our street photography.
Wow, these are beautiful, absolutely mesmerizing really.
Asking ChatGPT to Generate Own Art Prompts
But you can also allow Chat GPT to come up with its ideas.
So I’ve asked it to give me five varied ideas, including two that are photorealistic and one of a beautiful woman.
And we’ll take a look at the artwork that Chat GPT wanted to make.
What does the AI desire?
Hot air balloon scenes, wow, wonderful French countryside, and for the beautiful woman, it has a Japanese woman in traditional attire.
Interestingly, I would say perhaps there is an inbuilt bias there in Chat GPT to give us a very diverse output rather than a very perhaps narrow version of the most likely output.
And then we have a wonderful street scene.
So thank you to GPT for your contribution to the art here.
Bias for Attractive Women in Artworks
I love just scrolling through the Midjourney Community Feed and seeing what people come up with.
You would have to say that at least half of the images that are generated are based around attractive women, which is great because I love attractive women.
They can also ask Chat GPT for its opinion on why the majority of art created is based on attractive women.
Chat GPT has told us that the prevalence of attractive young women in AI-generated art can be attributed to several factors.
They reflect both the biases within the AI itself and the preferences of the users.
Should we address this?
I don’t think we should.
I don’t think there’s an issue with people choosing to create beautiful women, but it is something that happens.
If anything, it’s very interesting for me that on one hand,
I enjoy looking at attractive young women, but it seems almost fleetingly enjoyable.
It’s a bit like watching pornography, that it is fun, but it’s very hollow.
I think there is more depth, more nuance, and more beauty to be found in other places.
Creativity comes from combining tools in original ways, and using what we learned from this ChatGPT and Midjourney tutorial, we can create advanced, articulate prompts.
If you liked this ChatGPT and Midjourney tutorial, check out my AI courses below.