Generative AI Parameters

Refer to Swagger for full AI Generation API technical documentation.

Prompt Guide

Prompt Text

Knowing how to engineer prompts is instrumental to producing high quality images in generative AI. Prompts are composed of keywords and phrases that describe a subject and its styles. An effective text prompt may have several components, including:

• Main subject: the who of the prompt, e.g. Small puppy with a fluffy white tail wearing a red collar.

• Action: the what/how of the prompt, e.g. Joyfully carrying a long wooden stick.

• Surroundings: the when/where of the prompt, e.g. On a busy street corner at dusk. A small basketball court in the background.

• Visual aesthetics: how you want the images to look, e.g. Shot from above, soft yellow light, blurred background.

Examples

Prompt:

illustration of two people riding bikes in the distance of a desert
road. Cactus and rock formations on the sides of road. sun in sky

Prompt:

a building in the woods made of pink cotton candy, bountiful, all pink,
surreal, unusual, misty light, foggy, dusk or dawn, magic hour, pink
sky, magical, wispy, strange, fantasy

Prompt:

cloud burst of colorful powder suspended in the air over a black lava
landscape, ethereal, fantasy, whimsical, infinite, contemporary,
saturated, highly detailed, ultra realistic

Prompt:

studio portrait of a stylish, modern Indian woman with heavy shadows,
moody, colorful styling, atmospheric, filmic, film, proud, confident,
colorful backdrop

Additional Optional Parameters

Refer to Swagger for full AI Generation API technical documentation, including the list of accepted values for each parameter.

media_type: Specifies the type of content produced by the model.

aspect_ratio: Specifies the aspect ratio of the generated images.

mood: Specifies the lighting and other qualities of the generated.

seed: To create new images that have a similar aesthetic to previously generated iamges, include the seed value that was provided with that set of previously generated images. Seed can also be used to reproduce previous results by using the exact same prompt text and parameters as the intial generation, along with the seed value.