Stable Diffusion

Neil Haddley • August 1, 2023

Generating images from text descriptions using Stable Diffusion via Jupyter Notebook and Hugging Face — a deep learning text-to-image model

AIstable-diffusionimage-generationgenerative-aidiffusion-models
← Back to Posts

I used Stable Diffusion, a deep learning text-to-image model, to generate detailed images from text descriptions.

https://huggingface.co/CompVis/stable-diffusion

https://github.com/huggingface/diffusers

"stableenv" environment

I created an isolated Anaconda environment "stableenv" for my Stable Diffusion development.

I created the stableenv environment (adding jupyterlab, ipykernel and ipywidgets)conda create --name stableenv jupyterlab ipykernel ipywidgets

I created the stableenv environment (adding jupyterlab, ipykernel and ipywidgets)conda create --name stableenv jupyterlab ipykernel ipywidgets

I activated the stableenv environmentconda activate stableenv

I activated the stableenv environmentconda activate stableenv

I added jupyter labs support to the environmentipython kernel install --user --name=stableenv

I added jupyter labs support to the environmentipython kernel install --user --name=stableenv

I started the jupyter lab servicejupyter lab

I started the jupyter lab servicejupyter lab

I switched to the stableenv environment (kernel)

I switched to the stableenv environment (kernel)

I installed the stable diffusion pipeline dependencies (I could have used pip)!conda install -c conda-forge -y accelerate diffusers transformers

I installed the stable diffusion pipeline dependencies (I could have used pip)!conda install -c conda-forge -y accelerate diffusers transformers

I created a pipe instance using the DiffusionPipeline.from_pretrained(id) method call.

I created a pipe instance using the DiffusionPipeline.from_pretrained(id) method call.

The DiffusionPipeline.from_pretrained(id) method call returned a warning

The DiffusionPipeline.from_pretrained(id) method call returned a warning

I generated an image using the pipeline

I generated an image using the pipeline

In this configuration Stable Diffusion did not take advantage of the laptop's GPU.

In this configuration Stable Diffusion did not take advantage of the laptop's GPU.

In this configuration Stable Diffusion took 4 minutes and 19 seconds to generate an image.

In this configuration Stable Diffusion took 4 minutes and 19 seconds to generate an image.

I updated the pipeline instance to use the laptop's GPU

I updated the pipeline instance to use the laptop's GPU

In this configuration Stable Diffusion used the laptop's GPU.

In this configuration Stable Diffusion used the laptop's GPU.

In this configuration Stable Diffusion took 1 minute and 7 seconds to generate an image.

In this configuration Stable Diffusion took 1 minute and 7 seconds to generate an image.