Use HuggingFace Stable Diffusion Model to Generate Images from Text

I typed “Generate a picture illustrating AI for drawing a picture” to Bing’s Copilot. The picture above was generated by Copilot. Have you ever wondered how to build a model to generate pictures based on prompted text, like Copilot or DALL-E? Well, in this article, I will show you step-by-step how to use the Huggingface pre-trained stable diffusion model to generate images from text.

Install Huggingface Tansformers and diffusers
At the beginning of a notebook (I used Google Colab free version with T4 GPU runtime), type the following code to install the necessary libraries:

!pip install --upgrade diffusers transformers -q
Import the Necessary Libraries
import torch

from diffusers import StableDiffusionPipeline
from transformers import pipeline, set_seed
Set up an Attribute Class TTI
The TTI class specifies the HuggingFace model id, the generative model type, and some related attributes.