Generate Images
The Generate Images tool is the heart of Kenna AI, a class of tool called text-to-image. You can find it by clicking Workspace on the top right, and clicking Generate on the left bar. See it in action: Watch a video
The Generate Images tool will turn your words into images. The words that you use to describe that picture is called a Prompt. A prompt is a set of words that articulate all of the elements in an image: The type of image it is (photograph, illustration) the subject, the camera angle, the time of day, the location. You can write a full sentence, but succinct words and commas work best.
This is an example of a prompt. Notice that a prompt is the complete story, a total description of what is in the picture. When you revise a prompt, always describe the whole end result.
a realistic photo of men's leather shoes on the floor of a luxurious department store, brown shoe laces, shiny leather, high quality, bokeh background
A prompt is not like talking to a person, it isn’t “chatting”
Image prompts are more like describing what a garment looks like to someone that cannot see it, and if they misunderstand, you have to rephrase and choose your words more carefully, instead of continuing the conversation. For example:
CORRECT: “A deep red winter coat with a seam”
INCORRECT: “Add a seam and make it red”
The incorrect prompt is an incomplete idea. It is something said mid-chat, and does not describe the end result. Additionally, the word “coat” is missing. While that may be obvious to a human, the A.I. cannot understand if you no longer want to see coats at all, so it will not create one unless we are specific. Always visualize the final image, and describe that full image in a prompt.
Don’t worry about negative prompts yet, let’s try Positives first.
How to write a basic prompt
Choose Your Prompt.
Start by entering a prompt into the Positive Prompt field.
a deep red winter coat hanging on a levitating emerald coat hanger
How should we interpret this prompt?
Deep red coats are easy to come by, but how seriously should the AI interpret the words at the end of your prompt, how important are they really? There are two strategies to communicate what you want. You’ve already heard a little about Positives and Negatives, and the next is mastering Guidance.
Guidance (located below negative prompt) tries to answer this by assigning a value to how creative the A.I. can interpret your words. Adjust the guidance slider to control how closely the generated image will match your prompt. The higher the guidance is, the sharper the image may appear as the A.I. has higher confidence. A low guidance may produce more creative results, but may look less crisp or dreamy. A safe middle ground is a Guidance of 7.
Adjust the Guidance Slider
To interpret your prompt very literally, slide it more towards “strict”. If you want to give Kenna more creative freedom, move the slider towards “Creative”. While it may be tempting to keep the Guidance slider at maximum to make Kenna AI do exactly what you say every time, it may not have enough specific information on that topic, and high guidance can cause artifacts. Reducing guidance will fix it.
Select Aspect Ratio and Number of Images
Finally, select the aspect ratio for your images (such as square, landscape, or portrait), and choose how many images you want to generate (2,4,6, or 8.
Once you’ve made your selections, click “Create” and watch your creations come to life!
Video Demonstration
In this video, we will demonstrate writing prompts and generating images with the generate tool.
What to do when a prompt isn’t quite right
Almost everything is possible, we just have to learn how to phrase it in a way that the A.I. can better understand. This is called Prompt Engineering and it is a unique skill that is growing in demand.
Learning how to balance both weight and order of your words to make the KENNA AI takes a little practice. Watch the video again and pay attention to the Negative Prompt — elements that we don’t want to see. Here we can make subtle adjustments for context to steer the AI away from any bias in its training data. In the video, we are discouraging the likelihood of navy and blue shoes.
To learn more about prompt engineering, check our prompting advice guide
Additional Notes
Limited knowledge before 2021
Kenna AI was trained in part with billions of images from the public internet before 2021. As such, some pop culture references may not yet be understood. Please discuss and request any blind spots in our Discord community.
The Generate tool is different from a chat bot
Unlike a tool like Chat GPT, when we are using the Generate Tool, we are not “chatting” with Kenna AI. Every image is a fresh start, so write the full image description every time.
The Generate Tool does not understand prompts like “make a change” or “what if it was like this” commands, , because it isn’t talking at all. Generate is only creating what is in the prompt field.
When it’s time to make an edit, rephrase the original prompt or for spot changes try the Variations and Generative Fill tools instead.