MLX meets Stable Diffusion: Generate MacOS Native Diffusion Images!

Key Takeaways

this is amazing do you want to use mlx framework and run stable diffusion in your Apple silicon machine do you want to create a UI like this so that you can natively generate stable diffusion image from prompts that's exactly what we're going to see today let's get [Music] started hi everyone I'm really excited to show you about using stable diffusion in mlx I'm going to take you through step by step on how to create images using stable diffusion and interface like this but before that I regularly create videos in regards to Artificial Intelligence on my YouTube channel so do subscribe and stay tuned I've already covered mlx in detail step-by-step guide and also installed a local LOD language model in a complete different video which I will link that in the description below in this we are going to use table diffusion to create images from prompt as a first step cond create hyphen in mlx Python 3.11 and click enter next cond activate mlx and click enter next git clone ml explore mlx examples and click enter next navigate to the stable undor diffusion folder inside that pip install hyphen requirements.txt and click enter that will install all the required packages next pip install gradio because we are going to use gradio to create the user interface and click enter now our environment is ready now we can try a test prompt if you see the list of files this is the file which we are going to try and test text to image. Pi this will automatically generate image for you so here is the prompt python text to image. Pi a photo of an astronaut writing a hor on mass and the number of images to generate is four and the number of rows is two now I'm going to click enter this is for generating images without user interface this will automatically download the required models and then start generating you can see that it's done downloading now the image got generated it saved in out.png I'm going to click that and here is the image that is nice so if you see the text to image. py file this is how it looks it gets the prompt from the terminal and initiate stable diffusion and finally it saves it in the file we going to convert this to a grader user interface create a file called ui. Pi in that I'm modifying the text to image file with the generate image function and adding all the code within that function and finally it will return a image from array at the top I have imported grader as gr now I'm going to add the interface iFace equals gr. interface inside that I'm going to add the generate image as a function a text box with different sliders and then output as a image stable diffusion image generator as a title next if face. launch now the application is ready to run going to our terminal type python ui. pi and then click enter this will automatically open this URL so I'm going to click this URL and this is the user interface now I can type man in Mars need four Images number of rows one and click submit now it's generating those images and here is the image ready man in Moss that's nice I'm going to create again cat in Moss with totally two rows that means it will be created in two rows going to click Summit and here we got the image of cat in Moss in two rows and also you can see the gpus is peing during the time of image generation that's it now we have integrated mlx with stable diffusion as simple as that I'm going to create more videos similar to this so stay tuned I hope you like this video do like share and subscribe and thanks for watching