Midjourney but FREE! the EASIEST way to use Stable Diffusion XL AI Art Generator (Fooocus SDXL 1.0)

Stable diffusion is an amazing and free AI art generator, but for new users it can be very intimidating and confusing given all the different models, user interfaces, settings, optimizations, extensions, workflows and more that you can choose from and tweak to generate a better than average AI art for free. Automatic1111 is the most popular way to run stable diffusion locally and is relatively simple to set up, but it might not give you great outputs just from the prompt out of the box. Comfy UI is great, especially for SDXL but can get pretty confusing and messy. This is where fooocus comes in. While there have been a ton of new stable diffusion UIs, Fooocus has the street cred of being developed by the author and developer of control net a very useful and popular technique used by most intermediate and advanced diffusion users. The primary goal The focus is to allow you to generate great images just from the prompt without having to worry about settings and optimizations similar to mid journey while still being offline, open source and free. Okay, let's actually install and run fooocus. First, go to the GitHub repository of fooocus that I■ll link below. Then scroll down to download and under Windows, click on the click here to download hyperlink. This will start downloading the compressed file. After it has been downloaded please extract the 7zip file. Now, after you go inside the extracted folder, all you need to do is double click the run.bat file. If this is the first time you're running it, it will download and install all the required dependencies, including as the SDXL base and refiner models. The next time you double click run.bat it will just start the web UI unless there are updates. If you are an existing SDXL user and already have SDXL models downloaded for Auto1111 or some other UI, right after extracting the 7zip file and before double clicking the run.bat file copy the already downloaded models inside the focus/models/checkpoints folder. You can also add other SDXL fine tuned models here if you'd like. Now if you double click run.bat it■ll downl and install all the required dependencies except the models. Okay. Here's the web UI that will be launched by the run.bat file. Let's generate the image in the thumbnail. Here■s the prompt as well as the image generated using Midjourney V5.2. I'll just copy the prompt and hit generate. After around a minute it generates these two images. Pretty decent faces, the hands not so much. A completely different style than Midjourney. Let's try to change stuff by clicking on advanced. You can change the aspect ratios if you'd like for more landscape or portrait images. Image number is the batch size so since it's two, we get two outputs. You can also add a negative prompt if you'd like. Let's change the art style to fantasy art and see the results. Pretty great results, except for maybe the hands and the watermarks. These issues can both be fixed by LoRAs and Vae but slightly disappointing out of the box. Let's generate only one square image in anime style of a Freedom Gundam. Here's the result. Let's see a more cinematic interpretation of it. Finally, let's generate a background, first in cinematic style. Looks pretty good, but not exactly what I'm looking for. So let's actually remove any art style influence by selecting none. This looks super interesting. Let's try the RPG style on it. Looks super cute and comfy. Here are two more useful settings. Let's change from speed to quality. This will take a lot longer to generate since it basically doubles the number of sampling steps. You can also change the base or refiner model if you'd like. Focus also supports adding multiple SDXL LoRAs. Okay. Here are the outputs. I don't see much difference in quality between speed and quality, so it will probably depend a lot on the models and prompts. Please do experiment. Now a Fooocus specific setting is under advanced, called sharpness. So a lot of users have complained that SDXL by default doesn't produce sharp edges, but rather somewhat fuzzy output, especially for faces and so. Fooocus adds a sharpness of 2 by default, so you might have already noticed a difference when comparing the outputs to auto1111 or comfy. And that's it. I'll see you in the next vid. bye.

Channel: CoderX

Views: 1,340

Keywords: stable diffusion, sdxl, sdxl 1.0, stable diffusion tutorial, midjourney, free midjourney, stop using midjourney, free ai art generator, fooocus, fooocus stable diffusion, automatic1111, comfyui, stable diffusion xl, how to install stable diffusion, how to install sdxl, midjourney free alternatives, goodbye midjourney, free ai, free ai art, best ai image generator, BEST FREE alternative to Midjourney

Id: Dhbmmtz0meY

Length: 4min 46sec (286 seconds)

Published: Wed Aug 16 2023