Writing good prompts

AI takes some getting used to. So we’ve provided some quick tips below which we recommend you read.

You may think you want a picture of a cat. But are you looking for a photo, or a painting? Is it just the cat’s face, or the whole thing. From the front, or the back? You’ll need to provide this context to get the most out of your renders.

Quick Start Guide

In this mode, you provide a written description (known as a “prompt”) of the image you wish to render, which the AI produces.

Generally, longer (30+ words) and more precise prompts give the best results.

For example:

What's the secret to an excellent prompt?

It's the text and keywords that you provde which allow the AI to understand what you're trying to render.

The catch? Your words might not always be the best ones for the AI.

Here's a few steps we can take to optimise our prompts:

1) Prompts have an order

Prompts have an order - the earlier words are in your prompt, the more important they are. For example, in the following prompt you're more likely to get a brown dog than a black dog.

2) Apply appropriate emphasis

Make use of adverbs to apply emphasis or add detail to your prompt.

Will render differently do:

3) Be wary of being overspecific - short prompts often work great too

Context is inferred from all keywords in a prompt. For example, in the following prompt:

The word England may already infer "green grass", "beauty" and "sunshine", meaning there's no need for a longer prompt like:

4) Manually adjust weighting with () and []

Put parts of your prompt requiring increases empasis inside round brackets, and parts requiring less empasis in square brackets. Extending our previous example, the following prompt is likely to render a brown dog:

However, it can easily be adjusted to put less weight onto the "brown dog" and more onto the "the dog is black":

5) Try rendering multiple times

It's a good idea to generate multiple images for any given prompt - often they'll be very different. Just click the "Generate" button multiple times to do this.

6) Make use of Styles

Avolo comes with pre-built styles in the sidebar (more about these below). Want a different style though? You can easily add your own - just tell the AI the style you want. For example by adding "Victorian style" into your prompt.

7) Get to know the styles of famous artists

Avolo can generate art in the style of any well-known artist - and can even merge styles or artists together. Looking for something different? Try combining a few styles together in your prompt, for example by adding to the end of your prompt:

Using Avolo - Styles and Prompts

Avolo helps you build your prompt using selectable Mediums/Formats, Styles, Colour Schemes, and Lens Perspective. When applied, Avolo adjusts your prompt to provide optimised wording for each of these behind the scenes.

These can be disabled simply by not selecting a Medium/Format or Style, leaving the Colour Intensity set to minimum, and leaving Lens Perspective at 100mm.

Avolo also provides Prompt Enhancement by default, which may optimise your prompt behind the scenes. This can be controlled in Settings.

Using Avolo - AI Controls

The most important parameter you'll set is "Detail". This is how many "steps" Avolo should take whilst rendering your image. A higher number of steps means a render will take longer. It also has a habit of increasing image contrast. Generally:

<5 - you won't have an image yet

6-20 - low qualtiy

21-35 - medium quality

36-45 - excellent quality

45+ - generally not required

Advanced Mode

Advance mode enables many additional settings:

1) Wider range of "Detail Steps" - now choose between 1-100 instead of 20-70

2) Control of "Degrees Of Freedom" - this tells Avolo how tightly it should stick to your prompt. The higher the number, the less variance from your prompt that will be allowed.

3) Control of "Seeds" - Seeds are the random numbers that initialise Avolo for each image. If you run a prompt with different seeds, you'll get different results each time, whereas if you run the a prompt multiple times with the same seed, the rendered image will be the same. It's possible to use _Seed_ to control image style - creating consistent image styles across multiple prompts.

Finally...

You’ll discover there are some topics the AI is not good at rendering - for example faces. Creative workarounds often exist, but on occasion you may need to try a very different prompt or, in the worst case, look elsewhere for your image.

And remember, the AI takes your input as _guidance_ only - sometimes the result won’t be as expected, and it's often a good idea to generate multiple images for a single prompt - you can do this just by pressing the \"Generate\" button multiple times.