Recently, many fellow AI community followers and AI artists had asked me on Linkedin and AI art discussion forums how could I make Chat GPT command Midjourney with perfect prompts.
Well, the most important thing that you have to do to make ChatGPT help you in creating exact prompting is … make friendship with it. Yes, you heard it right. Not literally friendship as the bot doesn’t have a soul, lolz but make him understand the prompt formula very minutely.
Here is what I did.
Teach ChatGPT the prompt framework
Step 1: Log in to ChatGPT and then write the following in the prompt space in ChatGPT.
Here’s a Midjourney formula for you to understand.
‘/imagine prompt: [Image we are prompting], [5 descriptive keywords], [Camera type], [Camera lens type], [time of the day], [focal distance], [style of photograph], [by photographer name] –ar x:y Reply ‘yes’ if you understood the formula.
You can replace the Camera type with Art or illustration type. Instead of Camera lens type, you can use brush type, you can skip the time of the day, and other stuff based on the prompt you want to achieve.
This is what you will get as a reply.
“Yes, I understood the formula. It appears to be a formula for imagining a photograph or image, which includes a prompt for the subject matter, descriptive keywords, camera and lens type, time of day, focal distance, style of photograph, and the photographer’s name. The “ar x:y” at the end might refer to the aspect ratio, indicating the dimensions of the image.”
Just imagine how intelligent it is that it understood the “ar” as an aspect ratio. Yes, I am not cooking up the story, hence I have added the screenshot here.
But one thing, as more and more you will use these prompt snippets, ChatGPT will start replying to you in simpler terms like “Yes, I understood the formula” and will wait for you to ask the next question/ step.
After Understanding and confirming that the ChatGPT understood my prompt, I wrote the following to it:
“So since you have understood the prompt formula, kindly write me 5 Midjourney prompts of portals opening to a new futuristic cities”
And, these are what it generated:
Sure, here are five Midjourney prompts for portals opening to a new futuristic city:
- /imagine prompt: A shimmering portal opens to reveal a towering metropolis of glass and steel, with gleaming skyscrapers stretching into the sky. Keywords: futuristic, sleek, urban, innovative, dynamic. Camera type: DSLR. Camera lens type: wide-angle. Time of day: twilight. Focal distance: 24mm. Style of photograph: architectural. By photographer name: Sarah Lee. –ar 16:9
- /imagine prompt: The portal swirls open to reveal a bustling city filled with flying cars, neon lights, and holographic advertisements. Keywords: futuristic, vibrant, technologically advanced, fast-paced, cosmopolitan. Camera type: Mirrorless. Camera lens type: telephoto. Time of day: nighttime. Focal distance: 100mm. Style of photograph: street photography. By photographer name: Michael Wong. –ar 3:2
- /imagine prompt: As the portal dissolves, a gleaming cityscape of chrome and glass comes into view, with intricate bridges and walkways connecting towering buildings. Keywords: futuristic, intricate, innovative, metallic, interconnected. Camera type: Smartphone. Camera lens type: ultra-wide. Time of day: morning. Focal distance: 12mm. Style of photograph: cityscape. By photographer name: Lisa Chen. –ar 19:9
- /imagine prompt: The portal opens onto a vast and sprawling cityscape, with towering green buildings and lush, verdant gardens spilling out onto the streets. Keywords: futuristic, sustainable, eco-friendly, lush, green. Camera type: Point-and-shoot. Camera lens type: standard. Time of day: midday. Focal distance: 50mm. Style of photograph: urban jungle. By photographer name: Juan Rodriguez. –ar 4:3
- /imagine prompt: As the portal dissipates, a stunning cityscape comes into view, with towering skyscrapers and vast expanses of water, all bathed in the warm glow of a setting sun. Keywords: futuristic, aquatic, modern, luxurious, serene. Camera type: DSLR. Camera lens type: zoom. Time of day: sunset. Focal distance: 70-200mm. Style of photograph: cityscape. By photographer name: Emily Kim. –ar 2:1
Understand one thing, the aspect ratios will be random. Use it or change it as per the version of Midjourney that you are using.
4. Aspect Ratio: In V4 we struggled for a higher aspect ratio for most of the time, later they supported 1:2 and 2:1 and later they lift the bar for 16:9 as well as 9:16, but in V5 we can use almost any aspect ratio and even go till 4:1, 5:1 ratios. That means even cinematic renditions of images, like anamorphic frame formats ( ratios 2.39:1 ) which are used in cinematic creations, are now possible in V5. –ar 5:1
5. Tile: In V4 they had removed the possibility of getting Tiled patterns for some reason that was available in the older version of Midjourney V3. Now with this version V5 alpha version, they have brought in the Tile pattern again. –tile
5. Tile: In V4 they had removed the possibility of getting Tiled patterns for some reason that was available in the older version of Midjourney V3. Now with this version V5 alpha version, they have brought in the Tile pattern again. –tile
Image Weights Now you can have full control of image weights that were earlier removed from V4 –iw 0 to 2
v5 is “less opinionated” (They did mention this in the official Release Document) What does that mean? This means that in V4 we had to only write smaller text prompts, and Midjourney used to do the heavy lifting.But in V5, we have to write more detailed prompts that give more details to Midjourney about what the lighting, style, or mood of the output image should look like. If we do not put anything, the default style will be Photorealism.
However, there is more room for control of the output generation through extensive text prompts. That means, there is more bond between the input text and the output image.
What Needs Improvement:
Midjourney has already released the statement that V5 is still in the Alpha state and is open to further improvements. However, in the current state, many creators feel that the way they used to prompt craft in V4 is not working the same way in V4.
Many have trained Chat GPT 3.5 to create Prompt for Midjourney V4 but they are not working in the current version. We have noticed that V5 as a platform is extremely plain Jane and if NOT prompted will create Photo realistic images by default. If we want it to generate artistic images or if we want more creative freedom, then we have to train it with extremely detailed prompts. This is one of the new challenges which creators have to face who have been suitably trained with version 4 of Midjourney.
The next big thing where V5 still lacks is Typography. Midjourney V5 still rolls out Garbled Texts when the character is more than one in a word. While, Adobe Firefly has already claimed to address the issue and they are the first in the line, we users of Midjourney have to wait and watch how fast Midjourney can role out the Typography support in their platform.
Will it be V5a or V5b or a complete new version of Midjourney, only time can tell. Till then we creators have to keep our fingers crossed. Well, did I already tell you that the crossed fingers have already been addressed by Midjourney in V5 version.
