Wouldn’t it would be great if we can use an AI program to generate images? As I have embarked on a journey to understand AI & machine learning, wanted to share a quick and fun project that you can run on your local system which can translate your language and generate images accordingly.
Recently stable diffusion announced its AI module’s public release to be used by developers around the world. In this post, we would be using the same distribution.
Step 1: Prerequisite
- Install Anaconda distribution from here
- Make sure you have Git installed as well
Step 2: Download the latest model weights
Visit this URL, you would need to sign up at huggingface to access these files.
Get the latest distribution as shown in the below image
Now navigate to the “Files & Versions” tab and download the latest ckpt distribution
Step 3: Download Git repository
Use this link to download this repository on your local system.
Step 4: Copy ckpt file
Make sure you are in folder Stable-Diffusion, run below command to create a folder.
mkdir -p models/ldm/stable-diffusion-v1/
Now, rename ckpt file downloaded in step 3 and place in above folder created.
Step 5: Activate the environment
Run below command in terminal
conda env create -f environment-mac.yaml
If you get error like below
`CondaValueError: prefix already exists: /opt/anaconda3/envs/ldm`
then run below command with –force parameter like
conda env create -f environment-mac.yaml --force
Now run the command
conda activate ldm
it could take some time depending on your system capacity and internet speed.
Step 5: load Machine learning models needed by stable diffusion
Step 6: Fun time
You can use the below command to generate images
python scripts/orig_scripts/txt2img.py --prompt "a photograph of an astronaut riding a horse" --plms
Sample images would be created in the outputs folder.
Few Sample Images created by AI
There are some issues for Mac users, make sure to keep reading this post for the latest updates.