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
Deep dive into manual controls with the Render & it’s many parameters
/render is very similar to /brew, as it is the manual way to brew an image. Render has many hidden parameters. Render supports many fine-tuning commands to help you get the exact image you want.
Let’s start with simple render:
/render a man with a shield
How is this different from /brew? With /brew, the AI will most certainly pick a realism style and concept and slip in words for good anatomy.
With a /render, PirateDiffusion not going to add anything extra words or concepts to make this photo beautiful. What you type is what you get, like how a DSLR camera’s manual mode works. It is bare to the metal, you’re on your own to make it do your bidding.
Think about the camera entering the scene. Describe the art from near to far. Imagine you are holding a camera and entering the picture you are trying to make. What your lens is encountering nearest, and first? What are you focusing on? What can go towards the end?
Imagine the AI is starting to tune you out little by little as you add more words. If you organize your prompt this way you’ll find yourself shouting big ((((emphasis)))) later in the prompt because the overall instructions are just that much clearer.
MORE POWER
Pro users of PirateDiffusion have this giant library of styles and AI models to choose from, so let’s add one using <brackets> like this. Take a sneak peek at all the available concepts and styles.
EXAMPLE 2: A COMPLEX REQUEST
One of our users was struggling to make “an image of an angel holding a shield” look the way he wanted. He prompted:
/render ((angel wings)) (halo above head) full body shot muscular masculine angel , determined look, ((holding giant shield)) <eldenring> [[halo video game]][master chief]
The user is making good use of negative and positive prompts to add emphasis, but why didn’t it work?
SOLUTION:
However, before you resort to fancy tactics, let’s talk about what makes a good AI prompt in the first place.
Let’s think about what the ideal image would look like. A shield would likely be the largest and atypical object to add to a typical angel image, so when we’re hoping for out of the ordinary things to happen, or when things don’t appear at all, we should move them towards the front of the prompt.
Let’s switch from /brew to /render. A good structure is something like:
/render ((main important thing)) then the general topic, secondary important thing, in a location, [[unacceptable words]] <conceptname>
So instead of:
/render ((angel wings)) (halo above head) full body shot muscular masculine angel , determined look, ((holding giant shield)) <eldenring> [[halo video game]][master chief]
a MORE DESCRIPTIVE way:
/render (((a giant detailed shield)) is being held towards the camera, held by a muscular angel with (detailed angel wings) with a determined look towards the camera, (((holding a giant shield))) <eldenring>
You can also add “/guidance” to guide the AI on how creative or literal to be with your request. Where 20 is high and 0 is low, though you can exceed these values between -50 to 50 though you run the risk of blue box faces.
Since we’re talking about videogames, adding a few high-quality words associated with video game screenshots like “unreal” and “octane render” can also help influence better results. Putting all together:
/render /guidance:10 (((a giant detailed shield)) is being held towards the camera, held by a muscular angel with (detailed angel wings) with a determined look towards the camera, (((holding a giant shield))), unreal engine, octane render, video game aesthetic <eldenring>
Word order really matters. Instead of focusing on what the picture is, write the prompt from the visual weight of objects.
Alas, we got exactly what we wanted, with a shorter prompt!
A framework for your prompts
This isn’t a hard rule, but we think this kind of prompt template is effective. Let’s also review some more parameters possible with render. This is a lot to take in so pace yourself and try one parameter at a time:
/render /seed:x /guidance:x /style:x /size:x /steps:x ((main important thing)) then the character, secondary important thing about the, in a location, [[unacceptable words]] <conceptname>
go with the flow:
- we’re going to do a manual /render,
- starting with this specific image idea (seed is like a picture’s ID#),
- and use this much freedom to do what I’m about to say (guidance),
- using this general style template after my prompt (style),
- and make the image this aspect ratio and resolution (size),
- in this many passes for fidelity (steps),
- so start by sketching the most important thing (((positive prompts)))
- then the subject follows with some details,
- at this specific background situation or location,
- but definitely don’t do these things [[negative prompts]]
- and now that you have the ballpark sketch of what I want, make it look like this aesthetic <concept>
This flow will probably evolve over time, but this framework is very effective because the system does weigh words more heavily towards the front of the prompt. And we need some important commands before the prompt. Word order matters for brew as well as render.
Get more image variations
Select a photo, and reply to it (like you’re talking to the photo). On mobile, it’s a long press on the photo. On PC, it’s right click.
then type /more, like this:
/more
By default, the more parameter effect is mild. In the case of this portrait of a woman, it gave me subtle changes in hair, clothing, and facial expression.
This is useful when you like the image but something minor is a little off, or you’d like a slightly different facial expression, or you want the eyes to change slightly. If you look closely at her mouth, each image has a slightly different smile. So if you want something more different, we have to learn new ways to ask for that.
For beginners, skip to the next guide: editing images with AI
bonus: OPTIONAL SUPER NERDY GUIDE ON GUIDANCE AND STRENGTH PARAMETERS
If you’re insane, not in a hurry, or are on a mission to becoming a professional at this, you can go down this bonus rabbit hole and learn about the hidden parameters called /guidance and /strength.
Order right now to get instant access to all the PRO features
- Trusted by thousands of creators
- Easy to use, fun to master
- No GPU required
- Works on any phone or PC
- Creative privately, or join a group
- Royalty free
- Use it solo or with a group
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
Psst: Want the recipe for three elephants in a pub? Learn how to use Real Vision
Render-Ahoy!