Tips on Crafting Prompts

How to Craft Prompts for Image Generation

When it comes to creating compelling images with artificial intelligence, the secret lies in crafting effective prompts. The prompts you input guide the AI to generate the images you desire. Think of it like having a conversation with an artist: the clearer and more detailed your request, the closer their work will be to your vision.

Understanding Language Associations with Images

The AI that you're interacting with is like an intelligent art student who has studied countless pieces of artwork. It has been trained on a vast amount of visual and linguistic data, learning to link words with related images. For instance, when the AI hears "sunset," it understands to use warm colors, soft transitions, and the silhouette of the sun descending.

The AI interprets prompts based on the associations it has learned. However, it's important to remember that AI, unlike humans, doesn't truly understand the meaning behind the words. Instead, it leverages patterns found in its training data to produce a relevant response.

Simple steps to crafting Excellent Prompts

To create the most accurate images, you need to provide the AI with detailed prompts. Just like an artist, the AI benefits from clear instructions about your desired image.

Other image generators often don’t pass your prompt directly to a model - They interpret it for you, passing in their opinion on what quality outputs look like. Invoke requires more complete descriptions of what you’re looking for in order to produce quality outputs - We typically recommend including at least one or more terms in each of these categories:

Subject: This is the main theme or idea of your prompt, and typically should come early in the prompt. It could be a person, event, concept, etc. This will guide the direction of the response. In a drawing prompt, it could be "a mountain landscape", "a bustling city scene", or "a simply dressed villager".

Style: The style refers to the manner and medium in which you want the prompt to be executed. In a visual art context, this could refer to a particular artistic style (e.g., realism, abstract, impressionism) or medium (e.g., oil painting, pencil sketch, digital art).

Quality Modifier: The quality modifier specifies a particular aspect or trait that you would like to see emphasized or done exceptionally well in the response. This could refer to elements like detail, creativity, emotional impact, or accuracy. In the context of visual arts, the Quality Modifier refers to a specific characteristic, trait, or description that indicates a high level of quality or prestige in the resultant artwork. It might be a term or phrase associated with renowned or highly esteemed works, such as "award-winning", "featured in {relevant set of high quality works}", "professionally acclaimed", or "masterpiece". This descriptor provides further nuance on what would constitute a successful response to your prompt, leaning towards recognized high standards.

Aesthetics & Composition: Aesthetics & Composition relates to the overall visual impact and design of the artwork. It may refer to the color scheme used ("bright and vibrant colors", "cool tones"), the arrangement of elements within the piece, the use of light and shadow ("dramatic lighting"), the balance and proportion, or any other visual elements that contribute to the final look and feel of the piece. This provides direction for how the artwork should be visually arranged and the kind of sensory impression it should leave on the viewer.

An example of a detailed prompt might be: "A detailed oil painting of a tranquil forest at sunset with vibrant colors and soft, golden light filtering through the trees."

Model-Dependent Prompts

Finally, remember that the effectiveness of your prompts may depend on the specific AI model you're using. Different models have been trained on different datasets and may interpret prompts in slightly varying ways. Therefore, what works well for one model might not work as well for another. It's a bit like working with different artists who each have their unique styles and interpretations.

Don't be afraid to experiment and adjust your prompts until you get the results you desire. The AI models are robust and can handle a wide array of prompts, so feel free to get creative!

Negative Prompts

A negative prompt, also known as an unconditioning prompt, is used to express terms, elements, or conditions that are undesired or should be avoided in the final output. In the realm of visual arts, this could include characteristics that you don't want to see in the image such as certain colors, themes, styles, or subjects. These prompts provide explicit guidance on what not to include in the creation process.

Note, however, that this is an extremely strong effect. Using the negative prompt "moss" may inadvertently result in similar concepts like "grass" and "green" to also have diminished presence in the output.

Using Prompt Templates

Using a default prompt template is another great way to learn how to craft effective prompt language You can read more about using prompt templates here.

Prompt Examples

a friendly squirrel with red hat, Eco-conscious mobile game, vibrant kid-friendly graphics, green tech monthly visual appeal, whimsical nature-themed characters, colorful elements
Rustic log cabin, woodland setting, rustic living aesthetics, cool toned harmony with the surroundings, sustainable materials, integrated solar roof design
Urban nightlife, black-and-white, Hasselblad camera, international award winning, competition finalist, dramatic play of lights and shadows, sharp cityscape contrasts
Avant-garde couture, tactile textures, vogue aesthetics, vibrant color palette, intricate embroidery details, dramatic silhouettes
Abstract expressionist painting, sunflowers, bold brushwork, gallery exhibition quality, bright and vibrant color splashes, emotive visual narrative