# Inference steps

Inference steps in text-to-image diffusion models refer to the number of times that the model iterates over the diffusion process to generate an image. The more inference steps that are used, the higher the quality of the generated image will be, but the longer it will take to generate the image.

The number of inference steps that are required to generate a high-quality image will vary depending on the model architecture, the text prompt, and the desired level of detail. For example, generating a simple image from a text prompt, such as "a black cat sitting on a red couch" may only require a few dozen inference steps. However, generating a more complex image, such as "a realistic portrait of Albert Einstein" may require hundreds or even thousands of inference steps.

Some text-to-image diffusion models allow users to specify the number of inference steps that they want to use. However, other models automatically select the number of inference steps based on the desired level of detail.

Here are some tips for choosing the right number of inference steps for your needs:

- If you are generating a simple image, you can start with a lower number of inference steps and increase the number of inference steps if you are not satisfied with the quality of the generated image.
- If you are generating a complex image, you may need to use a higher number of inference steps to achieve the desired level of detail.
- If you are using a model that allows you to specify the number of inference steps, be aware that using a higher number of inference steps will increase the generation time.