GENERATIVE AI SYSTEM CARD

AI systems that generate images

Last updated Sep 27, 2023

One type of artificial intelligence (AI) system is known as generative AI because people can use it to create something new, like images and text, based on descriptions or questions provided as prompts. The information here focuses on the type of AI system that may be used by Meta to generate images, including those you may use on Meta’s technologies. Learn more about image generation on Messenger, Instagram and WhatsApp and on an AI’s profile on Facebook.

IN THIS CARD

How it works

Now, you try it

Overview

Usage tips

Data usage

What to be aware of

How it works

Meta’s AI systems that generate images typically rely on models that convert the words you provide as a prompt into an image. These models are trained by analyzing billions of images and their text captions (the descriptive text associated with the images). The model learns the association between these text descriptions and the images. After it’s learned the associations, it can generate new images when you enter a text description of the image you want to see.

There are multiple steps involved in this process, which are described below.

1

Prompt entry

The first step is for you to enter a prompt, for example, words that describe the topic you are interested in.

2

Safety mechanisms

Safety mechanisms analyze the prompt to detect harmful, offensive or inappropriate words that could produce problematic results.

3

Model processing

Next, a generative AI model converts your prompt into an image.

4

Result processing

Then, the initial image is transformed into a high-quality image through processing. This may involve removing noise, increasing the resolution, making color adjustments, or adding final touches to enhance the visual quality of the output.

At this stage, additional safety measures may be applied to reduce the chances of the AI system producing harmful results.

5

Image delivery

After the generated image passes the processing checks and undergoes modifications, the result is presented to you.

Now, you try it

Prompt the AI system to see what results you might get.

Send a prompt
First, you’ll need to prompt the system. This is typically done by sending it a few words describing the image you want to create.

Usage tips

There are things you can do to influence the responses you receive. Here are some tips:

Compose your prompt thoughtfully

Your prompt is the most important control you have over the result you will receive from the system. Try different prompts to see how they change the resulting image.

Use clear and specific prompts

For example, “Big white dog in sunglasses water skiing” will yield better results than “Dog skiing.”

Data usage

A large amount of data is required to teach effective generative AI models, so multiple sources are used for training. These sources include information that is publicly available online and licensed information, as well as information from Meta’s products and services. More details on how we use information from Meta’s products and services are available in our Privacy Policy and on the Privacy Center.

What to be aware of

Generative AI is still advancing, and there are important limits you should understand about how it works. For example:

Generative AI models can be used to create images that look real even though they’re not. To help prevent the confusion between images generated with AI with those that are not, we label the images that were created with our AI tools. However, images created with non-Meta technologies may not be labeled as AI generated. Learn more about our efforts to make images generated by AI easier to identify in Building Generative AI Responsibly.
The image-generation process is not consistent and as a result, the same prompt can generate different images.
Generative AI may produce images that are offensive. If you see something that concerns you, provide feedback in the app you’re using.

Product experiences