Beginner’s tutorial series
Easy mode with Brew

Positive/Negative prompts

AI upscale your images

Fine-tune with Render

Upload and edit images

> Change visual styles

Make realistic AI characters

Commands & troubleshooting

Change visual styles on the fly using the /style and <concept> commands

Pirate Diffusion arms you with two techniques for visual effects:

  1. Style templates (reusable text prompts)
  2. Focused image models, which we call <concepts>

But what’s the difference between the two?

A style template is simply text. Styles are created by you and me, when we have a good idea for a prompt that we want to use again. You will learn how to make and use styles by the end of this tutorial. You can use styles and concepts together to save time and create truly amazing images! Example:

/render /style:mycoolstyle    (Result:  Adds my favorite words)

A <concept> is “big data: aesthetic patterns learned from training.
Concept creators sit for countless hours, meticulously cataloging images with tags, and later “trains” a computer to reproduce those target patterns. Some concepts are created to do 1 thing well. Pirate Diffusion scours the web to bring you the top AI models preinstalled. The easy mode “brew” selects the right one for you, or go manual with render and pick one yourself. Example:

/render Obama, in a black suit of armor <realvis13>

Jump into the AI Concepts Directory!

We’ve been dying to show you this: Our concepts directory

You can browse our styles via Collegiate Style Spreadsheet of Doom ™ 

Why does the same prompt look so different across different models?

The short answer is different people made different models, and cataloged the images their own way. So there’s no way for them to all work the same way.

Some models are very broad in scope, while others are only good for making one specific thing, so this is very intentional by design.For example, a model like Anything V3.0 which was meticulously trained on anime.  So if you type /render  a girl <av3>, that will look quite a lot better than the limited anime data on /render anime girl (vanilla base model)

For example, look at these older Stable Diffusion models. These <concepts> will only output Balloons and Paper Cutouts. So when you try to create something specific, like a celebrity, pick the right model, or it may not understand who you’re taking about, while another model will nail it.

INTRODUCTION TO MAKING STYLES

Styles are good prompt ideas, generalized to be reused.

For example, let’s say that I discover that adding a negative prompt for [Zombie] is a great way to make beautiful people, and every time I want to make a beautiful person I’m always typing the same ten things… masterpiece, high res, 4K, etc etc.  You don’t have to do this every single time.

Instead, make a style:

/styles /new:MyBeauty $prompt
(((gorgeous, pretty, lovely, nice skin, shiny lips))) [zombie]

And now I have created a new style called beauty. The $prompt maker is where my text will appear in the future. I can now use it, like this:

/render /style:MyBeauty  Shrek looks at himself in the mirror

/render /style:MyBeauty  A donkey puts on lipstick

RESULTS: IT AUTO TYPES FOR YOU

…Rendering “Shrek looks at himself in the mirror
(((gorgeous, pretty, lovely, nice skin, shiny lips))) [zombie] 

…Rendering “A donkey puts on lipstick” [zombie]
(((gorgeous, pretty, lovely, nice skin, shiny lips))) [zombie] 

There are hundreds of such styles thoughtfully created by the Pirate Diffusion community that you can edit and make your own. Styles are intended to be personalized in your own unique way, and are not global (meaning each room has its own styles, and your private room has zero to begin with). So if you want styles in your private chat room, copy some over.

To learn more, type the /styles command in Telegram.  Remember, styles vary by room.

concepts: THE POWERFUL ONE

It is expensive and time-consuming to create an art model that does everything well. Lucky for us, training a small set of images to result in one aesthetic is not. These mini-models are hyper targeted to doing their intended purpose well. There are models for anime, memes, porn, and so on. They make getting great results so easy.

To see what concepts are available in the room you’re in, type /concepts

To copy a concept, do /concept /show:TheName and you’ll find the copy code at the bottom of the screen.  Copy that into the desired room to install it.

While we work on expanding our guides, we have samples on our favorites:

ANALOG

Mimicking an analog camera, these images have a professional and hazy feel to them. Contains nudity.
Prompt tips and more samples   |    Get PRO

CYBERPUNK ANIME

Cyberpunk Anime aka <cpunk> is one of the visual styles in AnimeMaker PRO, created by DGSpitzer. You can create amazing characters that resemble futuristic anime, similar to the movies Ghost in the Shell, Akira, and Battle Angel Alita.
Cyberpunk Anime samples   |    Get PRO

HASSAN’S BLEND

HassansBlend was trained on pin-up and explicit adult photography, and uses “danboro tags” (like 1man, 1girl) making it capable of producing very specific images. It is one of the most realistic AI models available to the public.

More hassan’s blend samples   |    Get PRO

MODEL SHOOT

Model Shoot is similar to Analog, but with a more contemporary look.
Get PRO

PORTRAIT PLUS

Great at portraits, real as well as illustrated. Contains explicit content.
Get PRO

WEDNESDAY JENNY

Just for one, at the request of one of our Patrons, we’ve added the Jenny Ortega as Wednesday Addams concept. It is trained specifically on the film. Get PRO

ANYTHING V3.0 and 4.0 – ANIME HD

High-quality anime AI model – Incredibly detailed. We have versions V3 with better VAE as well as Anything V4, and many other top anime AI models. Contains very explicit content, so the demo version has restricted keywords. Get PRO

Also see:

Prompt template styles – mix these with concepts for dramatic effects

Limitations

You can’t use multiple <concepts> at once, at least not yet. It’s like of like trying to wear multiple shirts at once, the top one basically wins and your prompt just looks somewhat fatter.

You also can’t use multiple styles at once, but there’s no limit to what goes into a style. So you can include quite a bit of tags from multiple styles you like and make your own frankenstyles that do what you need.

Beginner’s tutorial series
Easy mode with Brew

Positive/Negative prompts

AI upscale your images

Fine-tune with Render

Upload and edit images

> Change visual styles

Make realistic AI characters

Commands & troubleshooting

pirate diffusion logo

Render-Ahoy!