Writing a good prompt is both an art and a science. Experimentation, iteration, and practice are key to achieving success. Whenever you are crafting a new prompt ask the following questions:
Are you interested in having a photo or a painting made?
What would you like the subject of the piece to be? A person, an animal, or a landscape?
What special touches would you like to include?
Lighting: Soft, ambient, ring light, neon
Environment: Indoor, outdoor, underwater, in space
Color Scheme: Vibrant, dark, pastel
Point of View: Front, overhead, side
Background: Solid color, nebula, forest
Would you prefer a specific art style, such as 3D render, Studio Ghibli, movie poster, or something else entirely?
Are you interested in a particular photo type, like macro or telephoto?
Example:
Prompt: "A painting of a panda bear, neon lighting, in space, vibrant color, front view, nebula background"
The order of words matters
Let's see an example where word order changes the image:
Prompt: "A painting of a cute panda bear wearing a suit, natural light, in the sky, with bright colors, by Studio Ghibli"
Notice how we got the panda bear, the suit, the style, except he's not in the sky! The order of words in the prompt will have a big impact on the final result. Let's see if we can get what we expect by moving the words “in the sky” to the beginning:
Prompt: "A painting of a cute panda bear wearing a suit in the sky, natural light, with bright colors, by Studio Ghibli"
That's more like it! Remember, the earlier in the prompt a word or phrase is, the more weight it has.
Prompt Weighting
You can tell the model to emphasize and weight certain keywords higher by appending one or more plus '+' symbols to them. You can group tokens together using parenthesis '()'. Example:
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, sun rays"
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, (sun rays)+"
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, (sun rays)++"
Notice how the sun rays begin to dominate the scene even though they are last in the prompt ordering. Similarly you can de-emphasize tokens by using the minus '-' operator:
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, sun rays"
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, (sun rays)-"
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, (sun rays)—"
In addition to the '+' and '-' operators, you can also append a number between 0 and 2 (1 is default) to have tighter control over the amount of weight a set of tokens have. A number between 0 and 1.0 decreases weight, while a number between 1.0 and 2.0 increases it:
Prompt: "A cute panda in the forest eating bamboo, afternoon, beautiful warm lighting, (sun rays)1.5"
A '+' or '-' operator without a number is equivalent to 1.1 and 0.9 respectively. Stacking multiple +'s and -'s is equivalent to multiplying the numbers together. For example, (sun rays)+++ is equivalent to (sun rays)1.1*1.1*1.1 = (sun rays)1.331.As you can see, prompt weighting gives you more flexibility and control over your prompts than would be possible with tokens alone.
Limitations
There is a limit of 5000 characters for the prompt input field. You can use alphanumeric characters, spaces, and punctuation, along with emojis. The prompt is not case sensitive.
Modifiers
Modifiers can change the style, format, and perspective of an image, enhancing its visual appeal. Experiment with specific words and phrases to unlock the potential of your AI-generated art. Here are some examples to help get you started (Tip: click on a modifier to copy it):