An Advanced Guide to Writing Prompts – How to train an AI


  1. Context – Give it a persona or tell it who to emulate
  2. Task definition – Set a task and use colons to be very specfic. Write up to 5 tasks. Start or end point. The journey or question. Characters involved. Background information.
  3. Iterate – Change things like language style formal vs informal. Humour or professional.


Step 1:

Type “You are a prompt generation robot. You need to gather information about the users goals, objectives, examples of the preferred output, and any other relevant contextual information…” (Essentially what this is doing is getting ChatGPT to help you prompt it *perfectly*) (sample post)

Step 2:

For this example, I want ChatGPT to generate me a business outline.

“I want you to generate me a business outline. The output should be an exceptional business outline. The product I’m writing this outline for is a newsletter called “The Rundown”

Step 3:

ChatGPT will respond with a series of questions.

The more detail you provide in your questions will increase the detail of your outline.

ChatGPT will return an in depth response based on the answers we fed it.

Keep in mind, I only used the business outline example.

You can use this trick to get ChatGPT to generate perfect answers for anything you want.


This guide provides tips and keywords to enhance image output using better prompts. In order to get AI to do what we want we need to give text in a structured and detailed way to get what we want. IN this article we will explore various AI platforms and how to create great prompts.

Mdjourney is text-to-image generation app that generates stunning images based on the text prompts provided. It is currently in beta and constantly adding new features to its AI engine (v3).

The key to creating vibrant and unique images on Midjourney is to be descriptive in your prompts. However, there are advanced options available to generate variations in the output.


This keyword refers to the application of a specific artistic style, such as art deco, to a particular image. In the case of a cityscape, the stylization will give the image a unique visual appearance, incorporating elements of the chosen style to create a distinctive look. (0-100)

Chaos level

This keyword refers to the level of randomness or unpredictability introduced into an abstract landscape. The higher the chaos level, the more abstract and unconventional the image will become, as opposed to a more structured, traditional landscape. (0-100)

Stylization level

This keyword is similar to “stylize,” but refers to the degree of stylization applied to an image. A lower stylization level will result in a more realistic representation, while a higher level will bring out the more artistic and stylized elements of the chosen style.


This keyword refers to the detail and clarity of an image, and is measured in terms of pixels. A higher resolution, such as 4K or 8K, will result in a clearer, more detailed image, while a lower resolution will result in a grainier, less detailed image. You can specify the output resolution with keywords like 8K, 4K, photorealistic, or by using standard settings like HD and quality.

Aspect ratio

This keyword refers to the proportion of an image, and is expressed as the relationship between its width and height. A standard aspect ratio is 16:9, while a custom aspect ratio, such as 3:2 or 4:3, allows for a more unique and specific proportion. You can choose from default square output or select a standard aspect ratio like cinematic or wall-paper.

Applying weights to image and word prompts

You can assign different weights to each image or word prompt to give more importance to one over the other. The syntax is as follows:

/imagine fire dragon –wimg 1 –wword 3

This assigns a weight of 1 to the image prompt and a weight of 3 to the word prompt, making the word prompt have more influence on the final image output.

Image as a prompt

If you want to get some outputs similar to an image style , ( when you want to get consistent outputs ) across multiple images, pass a URL of the image

/imagine [image-name .jpg skyscraper, apples

The image generated will take cues from both the seed image ( from the URL you passed ) and also the text prompts.

You can give multiple images as prompts

You can specify weights to the images ( see the next point )

Weights to Image prompt

If you want your output to be looking more like your prompt image ( see the previous point), give a higher weightage to that image (keyword : --iw <number>) .

/imagine image-name.jpg skyscraper --iw:3

Weights to text prompt

/imagine wild animals tiger::2 zebra::4 lions::1.5

Filtering Out Words

Sometimes, you might want to filter out certain words that you don’t want to be considered in the image generation process. To do this, you can use the following syntax:

/imagine fire dragon –fword “dragon”

This will filter out the word “dragon” and only consider the word “fire” in the generation process.


By using these advanced options, you can tailor your image outputs to your exact specifications, making your image generation process smoother, faster, and more accurate. Whether you want to experiment with different styles, change the level of abstraction, or modify the resolution, these advanced options will help you get the results you want with Midjourney.


  1. “Stylize a cityscape in the art deco style, with a resolution of 4K and an aspect ratio of 16:9.”
  2. “Create an abstract landscape with a chaos level of 75, in ultra photorealistic resolution and a 1:1 aspect ratio.”
  3. “Design a portrait of Frida Kahlo with a stylization level of 50 and a resolution of 8K, with a custom aspect ratio of 3:2.”
  4. “Generate a surrealist still life with a chaos setting of 100, in HD resolution and a 1:1 aspect ratio.”
  5. “Create a photorealistic depiction of a sunset with a stylization level of 10, in 4K resolution and a custom aspect ratio of 4:3.”


Leave a Reply

Your email address will not be published. Required fields are marked *