AI Art has stormed the world since the beginning of 2022. It has become an exceptional launchpad for countless world-renowned AI artwork as it popularises. While many people have had previous exposure to AI-generated content (AIGC), they still cannot take advantage of the full potential of AIGC and how to generate appealing AI Art through it.
We will use a popular app called Pixie as an example to demonstrate the complete process of quickly creating a Text-Based AI Art utilising the app's newly introduced AIGC function.
FIRST OF ALL, WHAT IS AIGC ART CREATION?
By definition, AI Art creation is the process of converting text-based descriptions into graphical images through the use of artificial intelligence. You can unleash your wildest imagination through words, and AI can make it come true. We will demonstrate several examples created by Pixie AIGC.
As you can see from the examples, the AI will analyse the Subject and its details, Scene (Context), and Style of your choice to generate an image. The process of AI art creation is like the casting of a spell. The creator needs to describe the scene with words, and the AI will analyse your description and "summon" the results based on your depiction.
STEP1, START THE PROMPT WITH THE SUBJECT WORD
Suppose you are newly beginning your AI art creation journey. We suggest you start practising your prompt input by clearly defining the Subject. (Such as "Sky", "Room", or "Bike") Make sure you always describe the Subject in short sentences since long paragraphs can distract the AI system's focus, generating images different from your intentions.
STEP 2, ADD MODIFIERS TO YOUR DESCRIPTION (PROMPT)
With the Subject in place, you can now provide more details (modifiers) to help AI understand what you want to put into the final product by adding Context, Image Details and Styles. Here is an example.
A pretty woman (Subject), delicate features directed gaze (Description of facial expression), Earrings (Details), concept art (Style), 8k, studio lighting (Image details)
STEP 3, CHOOSE APPROPRIATE IMAGE ASPECT RATIOS
Choosing the appropriate ratio is also an essential part. For different contents, selecting a suitable aspect ratio prevent AI from adding unnecessary details or leaving out important information when drawing. Here are some suggested aspect ratios:
a. For Portrait (Characters)
1:1 for upper body only
3:4 for full body description.
b. Landscape or Fictional, Fantasy Content
4:3 for landscape results
3:4 for portrait/vertical results
However, none of the above are absolute standards. An excellent AI art creator can use many modifiers to control the results carefully, and AI has a certain randomness. It takes many attempts to output genuinely excellent work.
MORE ADVANCED TECHNIQUES
In the creation process, we found that AI is not as "compliant" with your description as intended, and there will be problems such as generation not matching the description and reparation. The reason is that the AI system may not understand the weight of different words, resulting in random generations. Here are some ways to fix the problem.
a. Adjust the order of words for a smoother AI analyse
b. Split long sentences into concise phrases and use punctuations to help AI understand the weighting between words
c. Avoid using a word that contains different meanings
d. Add more detailed descriptions (modifiers) for newly introduced Subjects, Scenes/Context, and Styles to increase their weighting
e. Avoid using repeated or contradicting modifiers in your description
In the actual AIGC creation process, even if you use a set of proven excellent descriptions created by others. There is still the possibility of unsatisfactory AI output. The solution is also straightforward, trial and error.