October 18

Mastering Midjourney: Tips for Making Prompts

What Is a Prompt?

A prompt is a brief description that Midjourney uses as a foundation to generate an image. A well-crafted prompt can help bring your vision to life. You can easily use Midjourney through @GPT4Telegrambot.

Prompts begin with the /imagine command and can include text, image links, and parameters:

  • The text describes what image you want to generate;
  • Image URLs can be added to a prompt to influence the style and content of the finished result. Image URLs always go at the front of a prompt;
  • Parameters change how an image is generated. They go at the end of the prompt with two dashes --
/imagine large letters "Midjourney" on a bright abstract background --ar 2:1 --style raw --v 6

How to Use Image URLs in Prompts?

You can use images as part of a prompt to influence a picture's composition, style, and colors. You can also combine two different images.

How it works:

  • Upload your image (or two) to Postimages.org and get links to them in .png, .jpg, or .webp format;
  • Paste the link at the front of a prompt and then add the text description;
  • To combine two images into one, use two links in a row separated with a comma.

Example of the prompt with image:

/imagine https://i.postimg.cc/MpDzn2xC/logo.jpg graffiti with a dragon on the wall

Example of combining images:

/imagine https://i.postimg.cc/HkSfGpTk/Sculpture.png, https://i.postimg.cc/xdcBnWdy/flowers.png

Parameter List --

Parameters are options added to a prompt that change how an image is generated. Parameters are always added to the end of a prompt. You can add multiple parameters to each prompt.

Here is the List of the Most Useful:

--v 5.2, or --v 6 To use an earlier version of Midjourney (version 5.2 or version 6, respectively). The default setting is Midjourney V6.1;

--niji and --niji 6 To produce anime-style images;

--aspect, or --ar changes the aspect ratio of the generated image. An aspect ratio is the width-to-height ratio of an image. It is typically expressed as two numbers separated by a colon, such as --ar 16:9 or --ar 4:3. The default aspect ratio is 1:1;

--style raw creates more natural and realistic images by reducing the influence of Midjourney's aesthetic processing;

--stylize or --s <number 0–1000> parameter also affects the artistic processing of the image. Lower stylization values produce more natural-looking pictures that closely match the prompt, while higher values produce more artistic images. The parameter ranges from --s 0 to --s 1000, with the default being 100;

--chaos <number 0–100> parameter influences how varied the initial image grids are. Higher values produce more diverse and unique initial image variations;

--weird <number 0–3000> introduces quirky and offbeat qualities to your generated images, resulting in unique and unexpected outcomes. The scale goes from 0 by default, indicating no "weirdness," to 3000, representing maximum "weirdness";

--no tells the bot what not to include in your image, --no plants would try to remove plants from the image.

Find out more parameters on the Midjourney website. @GPT4Telegrambot supports the following parameters: --v --niji --aspect --ar --style --s --stylize --chaos --weird --no --sref --sw --cref --cw --upbeta --seed --tile --uplight --hd --test --testp --iw.

Style References

Midjourney allows you to replicate the style of another image or maintain a consistent style across multiple pictures. For this, the parameter --sref.

How It Works:

  • Upload the image whose style you want to replicate to any photo hosting service, such as Postimages.org, and copy the URL in .jpeg or .png format.
  • When generating a new image, add --sref at the end of the prompt followed by a space and the link to the reference style image.
  • By using the additional parameter --sw ranging from 0 to 1000, you can control how strongly the reference influences the style. The default is --sw 100.
/imagine Iron Man --sref https://i.postimg.cc/RhQDgmP9/Spider.jpg

Style Random

Using the parameter --sref random allows you to assign a random artistic style to an image. It's an excellent tool for finding ideas and inspiration.

/imagine a fox --sref random

Character Reference

On Midjourney, you can generate images featuring the same character in different situations. This allows for creating stories, comics, and any content involving storytelling. For this, the parameter --cref.

How It Works:

  • Generate a sample image of your character, upload it to any photo hosting service like Postimages.org, and copy the URL in .jpeg or .png format.
  • When generating a new image, add the parameter --cref followed by the link to your character's image at the end of the prompt.
  • You can use the --cw parameter to modify reference strength from 100 to 0. By default, --cw 100 is applied, which copies facial features, hairstyle, and clothing. With --cw 0, it'll just focus on the face (good for changing outfits/hair, etc.).
/imagine a lady in a red dress is dancing --cref https://i.postimg.cc/sg006wJB/lady-in-black.jpg --cw 0

Note: This feature works best when using characters made from Midjourney images. It's not designed for real people/photos.

Prompt Length

Prompts can be very simple. Single words will produce an image. Short prompts rely heavily on Midjourney’s default style, so a more descriptive prompt is better for a unique look. However, super-long prompts aren’t always better. Concentrate on the main concepts you want to create. Remove words when possible. Fewer words mean each word has a more powerful influence.

Think About What Details Matter

Anything left unsaid may surprise you. Be as specific or vague as you want, but anything you leave out will be randomized. Being vague is a great way to get variety, but you may not get the specific details you want. Try to be clear about any context or details that are important to you.

Hint Commands for Prompt Design

Midjourney offers 2 helpful commands to assist you in creating a well-designed prompt:

/describe + image uploading. This command works in reverse: upload an image from your device's gallery, and add the /describe command in the caption. Midjourney will respond by sending several prompt options for that image.

/shorten and text prompt. Midjourney will analyze your prompt and suggest how to improve it: what words might not be doing anything and which might be key.

These commands do NOT consume generations, allowing you to look at prompts from Midjourney's perspective and better understand their design principles.

How to Use the U, V, 🔄, ✉️, and Zoom Buttons?

After sending the /imagine command with your prompt, the generation of 4 initial images begins, usually taking 1-3 minutes. Along with the images, two rows of buttons will appear:

  • U 1-4 – upscales the selected image by adding more detail;
  • V 1-4 – creates four new variations of the selected image;
  • 🔄 – regenerates the images using the same prompt again;
  • ✉️ – allows you to obtain the "seed" of the image. Midjourney uses a seed number to create a field of visual noise as a starting point to generate the initial image grids. Seed numbers are generated randomly for each image but can be specified with the --seed parameter. If you use the same seed number and prompt, you will get similar final images.

After upscaling (U) the initial image, new buttons appear:

  • Upscale – increases the resolution by two or four times. More details below;
  • 🪄 Vary – creates four new variations of the image with either significant (Strong) or minor (Subtle) changes;
  • 🔍 Zoom out (1.5x or 2x) – zooms out of the image and adds new details;
  • ⬅️ ➡️ 🔼 🔽 Pan – extends the image in the selected direction.

How to Get High-Resolution Images?

Upscale buttons allow for increasing the resolution. The buttons differ for Midjourney V6 and V5.

  • For Midjourney V6 and V6.1: Upscale (Subtle) and Upscale (Creative) both double the resolution to 2048x2048 pixels. "Subtle" aims to keep changes minimal, while "Creative" introduces more artistic alterations.
  • For Midjourney V 5.2 and earlier: Upscale X2 and X4 increase the resolution by two or four times, reaching 2048x2048 or 4096x4096 pixels, respectively.

The commands Upscale X2, "Subtle", and "Creative" consume two regular image generations, Upscale X4 consumes four generations.

How to Place Text on an Image?

Midjourney V6 can generate a short text on images: Write the text in English and enclose it in "quotes."

/imagine a photo of the text "Hello World" written with a marker on a sticky note

How to Make Images Photorealistic?

Midjourney generates incredibly photorealistic images that look as if they were taken with a camera or phone. How to achieve such realism?

  • Add the --style raw parameter, which produces more natural and realistic images;
  • Use lower stylization values, for example: --s 0;
  • Avoid clichés such as "4k", "8k", "photorealistic"; it doesn't work;
  • You can specify the type of camera or film used for the "shot." For example: "Shot on Canon R6 Mark II camera, 35 mm lens" or "Shot on iPhone";
  • Specifying the parameter --ar 9:16 will result in a vertical iPhone-style photo;
  • Midjourney V6.1 is the default setting.

Selection of "camera" and lighting on Midjourney >

Mastering Angles and Perspective >

Examples of Prompts

Examples are provided in English, Midjourney's primary language. However, @GPT4Telegrambot automatically translates your request so that you can write the prompt in any language.

Portrait of a young woman from Sicily

/imagine young woman from Sicily, portrait shoot, well-defined facial features, long black hair, photo taken with Canon R6 Mark II, 35 mm lens --style raw

The Terminator promotes the "Hi, AI!" channel

/imagine a robot-terminator is holding a yellow sign with the words "Hi, AI" written in black letters --style raw

Pomegranate seller, "photo taken on iPhone"

/imagine pomegranate seller at the Sicilian market --style raw --s 0 --ar 9:16

The race

/imagine photo shot of scene of a Formula 1 racing, fast speed, stadium, panoramic view, crowd

Elon Musk is riding a camel in Paris

/imagine Elon Musk is riding a camel in Paris, photo on a Canon R6 Mark II camera, 35 mm lens, uplight --style raw

Example of a cat image with the parameter --weird

/imagine cosmic cat --stylize 500 --weird 500

Are humans conscious?

/imagine the robot is sitting in an armchair and reading a newspaper with the inscription "Are humans conscious?"

The goblet in hands, close-up

/imagine detailed close-up image of a female hand, carefully holding an ornate glass goblet

Example of using an image in a prompt

/imagine https://i.postimg.cc/2ymHbJCt/screen.jpg cinematic, professional color correction, technopunk, science fiction, volumetric lighting

My dog Julia in Van Gogh's painting

/imagine a portrait of a poodle in Vincent van Gogh's painting "Starry Night"

A vast desert

/imagine a landscape photo of a vast desert, with a lone camel in the distance, overheard sunlight, midday --ar 3:2 --v 5.2

Sunset over the Great Wall of China, Impressionism

/imagine an impressionist painting of a sunset over the Great Wall of China

A hipster lion portrait

/imagine a hipster lion portrait, headshot --no glasses

Black poodle in the world of the game The Last of Us

/imagine a black poodle in the world of the game The Last of us --v 5.2

Queen of Fire with the --niji parameter

/imagine the queen of fire, the crown made of flames --niji

Portrait of a young woman with the parameter --niji 6

/imagine young woman, portrait, red t-shirt with "Niji 6" text, beautiful face, black hair --style raw --niji 6

Initial generation with a low level of chaos, --chaos 0

/imagine house by the water --chaos 0

Initial generation with a high level of chaos, --chaos 100

/imagine house by the water --chaos 100

Now, it is your turn to create a prompt and see what Midjourney can do for you ❤️🔥🍀

@GPT4Telegrambot