Artist Guide to AI Image Generation

What keywords should you use to better control ai image generation (for Dall-e-2 and MidJourney)

This will be a 2 parts article that will try to summarize the most useful keywords used for Midjourney and dalle. Let’s start with Midjourney. Most of the stuff can be found in their user manual found here: https://github.com/midjourney/docs/blob/main/user-manual.md

The format of the command is:

/imagine <IMAGE PROMPT> <TEXT PROMPT> <PARAMETERS>

For Image Prompts you can give an image link and take the essence of one or multiple images and distill it into a source for inspiration.

For Text Prompts the only limits are our language and imagination. An underused characteristic is the Weights. We can break our text prompt into parts and put a weight value on each part so that we can give emphasis to specific parts. If the weight is not specified, it defaults to 1.

Weights

/imagine oil paiting of a baseball field,::1.5 simple color pallet::1 trippy and dreamlike::0.25 green::-2 in the renaissance style --Test --Uplight --Video --Seed 1

Since we have specified green=-1 it means that our image will not have green at all:

Let’s undo the trippy part by setting it to TRIPPY AND DREAMLIKE::-0.25

Let’s also let it have green colors Green::1

Possible values for parameters are:

-- WIDTH, --HEIGHT, --AR, --UPLIGHT, --HD, --TEST, --FAST, --VIBE, --VIBEFAST, --STOP, --SEED, --SAMESEED, --VIDEO, --no, -iw
--AR

Aspect Ratio. Generates images with the desired aspect ratio. Try --ar 16:9 for example, to get a 16:9 aspect ratio (~448x256

--uplight 

This is a light version of upscaling. I think it's very useful to use for faces since it makes them less detailed and hence less error-prone.

--HD

It uses a different algorithm to add maximum details. It's like the opposite of the previous uplight. Best for abstract and landscape prompts. This also generates images at higher resolutions without the need to upscale. It is supposed to be better for larger images but with some sacrifice to form. (do not like)

--TEST

This is a model for “prettier” output (recommended)

--FAST

Runs generative model faster and gives more abstract results or painterly.

--VIBE

Great for textures and vibes but the details are lacking

--SAMESEED <value>

We can use it in order to generate the same (kind of) images in the grid of the 4 images)

Now the available styles are so many (there are currently over 1900 styles and keywords)! that you can find them here:

MidJourney-Styles-and-Keywords-Reference/Themes_and_Design_Styles.md at main ·…

or if you like you can check all (about 800) artists here:

https://docs.google.com/spreadsheets/d/10i9Ip8tVSERAuMWbc6-H6BUFCoUGOQ91YzDvX--c4bk/edit#gid=0

Lastly, I would like to mention here my favorite keywords:

ultra realistic, cinematic, reflective glowy, light neon, cyberpank, synthwave,realistic photo, cinematic composition, trending on artstation,hyper realistic photograph, real humanoid

Special Thanks to https://www.youtube.com/watch?v=IcNeyhfSoBQ for the very nice examples.

— — — — — —

Although DALL-e2 is a fantastic tool, you have to pay for more credits if you want to use in a project. By default, OpenAI gives you 15 credits(images) per month but you can buy 115 more credits for 15$. There is however a free alternative that uses the Stable Diffusion Model called DreamStudio. Accessing Dreamstudio is free and easy. It works exactly as DALLe2.

Following is a part of their FAQ guide which contains really good instructions on how to create interesting results.

DreamStudio

2. Style

Style is a crucial part of the prompt. The AI, when missing a specified style, usually chooses the one it has seen the most in related images, for example, if I generated landscape, it would probably generate realistic or oil painting looking images. Having a well chosen style + raw prompt is sometimes enough, as the style influences the image the most right after the raw prompt.

The most commonly used styles include:

Realistic

Oil painting

Pencil drawing

Concept art

I’ll examine them one by one to give an overview on how you might use these styles.

In the case of a realistic image, there are various ways of making it the style, most resulting in similar images. Here are some commonly used techniques of making the image realistic:

a photo of + raw prompt

a photograph of + raw prompt

raw prompt, hyperrealistic

raw prompt, realistic

You can of course combine these to get more and more realistic images.

To get oil painting you can just simply add “an oil painting of” to your prompt. This sometimes results in the image showing an oil painting in a frame, to fix this you can just re-run the prompt or use raw prompt + “oil painting”

To make a pencil drawing just simply add “a pencil drawing of” to your raw prompt or make your prompt raw prompt + “pencil drawing”.

The same applies to landscape art.

3. Artist

To make your style more specific, or the image more coherent, you can use artists’ names in your prompt. For instance, if you want a very abstract image, you can add “made by Pablo Picasso” or just simply, “Picasso”.

Below are lists of artists in different styles that you can use, but I always encourage you to search for different artists as it is a cool way of discovering new art.

Portrait

John Singer Sargent

Edgar Degas

Paul Cézanne

Jan van Eyck

Oil painting

Leonardo DaVinci

Vincent Van Gogh

Johannes Vermeer

Rembrandt

Pencil/Pen drawing

Albrecht Dürer

Leonardo da Vinci

Michelangelo

Jean-Auguste-Dominique Ingres

Landscape art

Thomas Moran

Claude Monet

Alfred Bierstadt

Frederic Edwin Church

Mixing the artists is highly encouraged, as it can lead to interesting-looking art.

4. Finishing touches

This is the part that some people take to extremes, leading to longer prompts than this article. Finishing touches are the final things that you add to your prompt to make it look like you want. For instance, if you want to make your image more artistic, add “trending on artstation”. If you want to add more realistic lighting add “Unreal Engine.” You can add anything you want, but here are some examples:

Highly detailed, surrealism, trending on art station, triadic color scheme, smooth, sharp focus, matte, elegant, the most beautiful image ever seen, illustration, digital paint, dark, gloomy, octane render, 8k, 4k, washed colors, sharp, dramatic lighting, beautiful, post processing, picture of the day, ambient lighting, epic composition