YandexART overview

YandexART is a generative neural network that creates images based on a text request. YandexART uses the cascaded diffusion method to iteratively elaborate the images from noise.

In addition to a text description, you need to specify a special parameter called seed to generate an image. A seed is any number from 0 to 2⁶³-1. It serves as a starting point for image generation from noise and allows to achieve repeatability. Thus, with the same prompt and seed, the generation result will be the same. To change the generated image, change the seed value or the description.

You can also specify the format of the final image in the mime_type parameter. Currently the supported value is image/jpeg.

Recommendations on creating prompts for YandexARTRecommendations on creating prompts for YandexART

To get a usable image, follow these tips:

1. Use vivid imagery, e.g., the following:
   • Winter forest
   • Sea sunset
   • Medieval castle
   • Futuristic car
   • Fall tree
2. Provide more details for objects, such as color, shape, location, and others.
3. Specify styles if you want to get a certain type of image. For example:
   • Aesthetic, beautiful, realistic, close-up.
   • Professional photo, 4k, high resolution, high detail.
   • Dali style, Monet style, Repin style.
   • Pixel graphics, Minecraft style, anime style.
   • Bright lighting, dim lighting, soft lighting, neon lighting.
4. Use descriptions instead of rare and complex concepts. For example, instead of using a little-known holiday name, you can write that people eat candies and stroll in the city square.