Stable Diffusion, Detailed Information on Prompts, Lighting, Rendering, Styles

What is stable diffusion?

An open-source text-to-image artificial intelligence application. This application is capable of generating stunning images from the given text in just a few seconds.

As you may know any artificial intelligence application requires a thorough training before you can put it to practise. Even stable diffusion has undergone thorough training on images in the LAION-5B dataset. Stable diffusion was developed by CompVis, Stability AI and RunwayML. All of the research knowledge and application from Stability AI are open source.

You might like these

In order to generate any picture, or art using Stable Diffusion, you need to feed a detailed description of the art/image that you have imagined or visualised. For e.g. the text could be:

Photo of a man from the 17th century using a laptop.

This text produced the below image on Stable Diffusion.

This text description which is fed to the AI application is called a “prompt”.

As a concept, prompt engineering is a natural language processing. Under this aspect of AI, you determine the input text that AI can process and provide you with a visual output. It basically means telling a description to the AI in a language that it can understand, process, and act upon.

Here’s another example of a prompt.

Funko pop superman figurine, made of plastic, product studio shot, on a white background diffused lighting, centered.

This prompt yielded the below result on Stable Diffusion.

Let us now understand what the prompt should include in order to get the right kind of result. To do this we can consider the below prompt and break it apart into different parameters. We have recently added a new article about how to generate realistic landscape photos using AI ART generator. You can also go through this article to understand how to write an AI art generator prompt.

Tiny cute giraffe using a typewriter toy, standing character, soft smooth lighting, soft pastel colors, by skottie young, 3D blender render, polycount, modular constructivism, pop surrealism

Break up of the AI ARTWORK Text Prompt:

Your subject is a TINY CUTE GIRAFFE.
What is the giraffe doing? USING A TYPEWRITER TOY WHILE STANDING
What kind of lighting should be used in the scene? SOFT SMOOTH LIGHTING
What kind of colors should be used in the scene? PASTEL COLORS.
If you have a specific kind of drawing or art in mind, you can mention the name of the artist so AI can replicate their style to produce the artwork you need. In this case the artist is SKOTTIE YOUNG.
If you have a specific style of render in mind then you can mention the style in the prompt. In this case it’s 3D BLENDER RENDER, POLYCOUNT, MODULAR CONSTRUCTIVISM, POP SURREALISM

All these details are mentioned comma separated and submitted in the input box of Stable Diffusion or any APP that uses Stable Diffusion or Dall-e (another AI art generator).

While your subject can be anything that you want, there are many options available to you to try for lighting, rendering, etc. Let us see what these options are.

Examples of lighting that can be used with Stable Diffusion are:

Bloom
Indirect light
Global illumination
God rays
Hard shadows
Studio lighting
Soft lighting
Diffused lighting
Backlighting, rim lighting
Volumetric lighting, specular lighting
Cinematic lighting
Luminescence
Translucency
Subsurface scattering
Radiant light rays
Bioluminescent details
Ektachrome
Glowing
Shimmering light
Halo
Iridescent
Caustics

Examples of Rendering methods that can be used with Stable Diffusion are:

Unreal Engine
Octane Render,
Bokeh
Raytracing
CGI
Lumen Reflections
CGSOciety
Ultra Realistic
100mm
Film Photography
DSLR
Cinema 4D
Studio Quality
Film Grain
Vray
Houdini Render
Quixel Megascans
Arnold Render
8K
UHD

In order to further improve on your AI-generated ART, you can mention specific styles too:

Analog
Polaroid
Macro Photography
Volumetric Fog
Depth of Field
Silhouette
Motion Lines
Motion Blur
Fisheye
Ultra-wide Angle
Mist
Overglaze

There are many other details that you can provide such as:

HQ i.e. high quality
Intricate, if you’re looking for intricate design elements.
Ornate
Crytalline
Perfectionism
Max Detail
4k UHD,
Spirals
Tendrils
Embroidery
Insanely Detailed
Lush Detail
Poster
Sharp Focus
Hyperrealism
Realist
Hard Edge
Breathtaking
Embroidery

What we recommend is that before actually trying to create your first proper artwork, you should get some practice with the above-mentioned parameters. For that, you can try using various permutations combinations, append your prompt, modify it and arrive at the perfect first prompt that will yield the desired AI Art result for you.

You can create an account on Midjourney AI, Playground AI, Dall-E, and practice. For practicing, the best website would be Playground AI as it provides you with the maximum image creations in a day as compared to others which do not offer too much for free.

Or if you have a high-end computer and a minimum of 4GB Nvidia Graphic Card, then you can install Stable Diffusion on your computer us the open source for free.

Happy AI Art Generating.

I recently cracked a near-perfect CHATGPT prompt that in return creates a prompt to generate stunning and realistic landscape photos. Check out how it works here.