Yes, you have come to the right place to learn the secret to consistently obtaining great photos with Midjourney. It's all thanks to a proven ChatGPT formula. So it's AI creating AI—with the help of a human Photographer, that is.
If you want only the formula, it's further down on this page. But if you have the time and want to understand what's behind this “secret sauce,” I recommend reading the whole article.
Make yourself a cup of tea, sit comfortably, and let's get into it.
A Brief Introduction
First of all, I hope that ChatGPT is still available in your country. As much as I understand the need to regulate new technologies, I disapprove of banning them altogether, as Italy once did. But this is not a point I want to make with this article (though I am happy to discuss it in the comments section below).
Of course, the formula still works without ChatGPT. But you will have to be creative with the inputs and fill in the blanks yourself. It's easy—it just takes a little more effort.
No, I'm not going to explain ChatGPT, Midjourney, Artificial Intelligence, image generation, etc. This would require several other blog posts. And I hope that if you landed on this page, it's because you have at least a basic understanding of what this is all about.
Here I will be referring to ChatGPT 3.5 (free) and Midjourney v5 (paid) exclusively. This is what I tested. I am confident you can obtain great results with other AI applications, but I have not tried the formula outside of this combo. If you do, please let everybody know your results in the comments section.
Also, I want to clarify that whatever you achieve with this ChatGPT formula is not a photo: it's an image. It's not just semantics. I hope you will not use the formula to try to fool everyone into thinking you have taken a photo. You can likely get away with it on social media and avoid the backlash of being spotted, but you won't land a job at National Geographic with it.
My goal here is simply to help you create beautiful images you can enjoy. Please refrain from using it to cheat.
Midjourney v5
The development of Midjourney has been moving way faster lately. It has taken much less time to go from v4 to v5 than it took between v3 and v4. And v6 has already been announced! The current version is a big leap in terms of photographic quality, most notably in terms of rendering hands and facial expressions. People were already obtaining outstanding results from the previous versions, but now, it can all be achieved way more consistently.
One main difference with the latest version is how you describe your desired image. With previous versions, it was all a series of keywords. With v5, Artificial Intelligence better understands a fully articulated scene description. This is very important to remember if you want Midjourney to return the image you want. And it's also the foundation of the ChatGPT formula you'll see in a moment.
So, if we take the image above, we go from “child running, wildflowers field, summer, soft light, octane render” to “a child gleefully running through a field of wildflowers during a warm summer afternoon”.
Huge difference.
More Details
What is even more important is that Midjourney now knows photography. The basics, at least. Kind of.
It could already output different images according to your lighting definition: backlight, direct sunlight, natural light, accent light, etc. And broad lens specifications, like wide-angle, pinhole or macro. But now you can go much deeper into the fine details.
What if I told you, you could ask Midjourney to create an image as if it would be a photo shot with a Canon EOS 5D Mark IV? Perhaps using a 24-70mm lens at f/2.8? This deeper.
Yes. You can.
And with the vast array of information available to ChatGPT (even with the limited free version), you can turn this into an endless stream of ideas.
Asking ChatGPT for Help
ChatGPT is really good at one thing, and that is conversation. I used it in different ways in these past months… For example, one article in this blog was written with ChatGPT. And I used it to create a plugin for WordPress in PHP because I needed something very specific for this blog. But I still needed to fact-check and edit the article. And the PHP code had to go through several iterations to eliminate the many errors, even for something quite simple.
But if you ask ChatGPT to talk about something, we can all agree it makes for a good conversation. And so if you ask it to describe something, it can surely come up with many ideas. It only requires you to point it in the right direction.
And so, for the ChatGPT formula to work, you need to specify what you need. You want to create an image that reproduces a photo. And a photo mainly shows light, a subject and an action.
The starting point is, therefore, obvious: a photograph of a subject performing an action in a context under a specific light.
The ChatGPT Formula
A photograph of a [subject, one or more] [engaged in an action scene] with [background context] during [time of day] with [type of lighting] and shot with a [type of camera and lens: brand, focal length and aperture] using [type of composition] and captured on a [type of film or film simulation]If you read through this post, you already saw the prompt in the images' captions. Now, allow me to elaborate… This expands on a couple different formulas already available online (either YouTube or Reddit), with the addition of technical details only a photographer would think of.
When I asked ChatGPT to let me know if it understood the formula, this is verbatim what the AI answered:
- [Subject] The main subject or subjects of the photograph, which could be anything from a person, animal, object, or scene.
- [Engaged in an action scene] Describes the main subject or subjects in action, engaged in some kind of activity or movement.
- [Background context] Refers to the surroundings or environment in which the main subject or subjects are placed, providing context to the scene.
- [Time of day] Specifies the time of day when the photograph is taken, such as morning, afternoon, evening, or night.
- [Type of lighting] Describes the lighting conditions in the photograph, such as natural lighting, artificial lighting, or a specific type of lighting, such as soft, harsh, or dramatic.
- [Type of lens] Refers to the type of lens used to capture the photograph, such as wide-angle, telephoto, prime, or zoom.
- [Type of composition] Specifies the composition technique used in the photograph, such as rule of thirds, leading lines, symmetry, or a specific composition style.
ChatGPT also understood that “each element in the formula is enclosed in square brackets [], indicating that it can be customized or varied to create different examples”. That is indeed correct; well done.
I used Novel AI's Tokenizer to double-check the input. While this formula seems to use way more tokens than the 72 that Midjourney accepts, it still produces better results than making the formula less verbose.
Expand on the Formula
You can, of course, leave the formula as is and let ChatGPT work its magic, but you can also narrow down its results to anything you want. Rather than a random subject, you may want to have a specific person or character in the image. To do that, you simply need to change a [subject, one or more] with whatever you fancy.
And the same goes for any other parameter or for more than just one.
Furthermore, last week, Midjourney introduced permutations. I tried to edit this formula to include such permutations and succeeded. However, it can easily lead to over 40 different prompts, and right now, permutations are only available to the “fast” paid tier, so they are too expensive to experiment with (as of yet).
But the formula is so good that you won't need to generate many images to get one you like. All the images I shared in this post were generated on the first try.
At its development pace, I'm sure Midjourney will introduce many more improvements that will make this formula even more effective. I will try to keep up and publish new updates whenever they seem significant enough.
But already, with this version, you have a lot to play with. I'm sure you will have a jolly good time with this ChatGPT formula for Midjourney.
I'm sure more AI models will enter the market soon. It's still early days and ChatGPT is pretty much the sole option right now.
So keep your eyes peeled for updates, and come to read about them. I will definitely update this formula when the time comes.
By the way, please mention this blog as your source if you re-use the formula or present it in your social media content… Let me know if I was able to help you!
