Fooocus v2: Free Midjourney Guide for Your PC

  • This topic is empty.
Viewing 0 reply threads
  • Author
    Posts
    • #17663
      imeda
      Participant

        Friends, hello!

        Today I want to tell you about the simplest and most understandable neural network that creates images based on your textual description. It’s called Fooocus and is based on the famous Stable Diffusion XL. It’s the perfect solution as your first neural network, and an essential tool for any designer or content creator.
        <p style=”text-align: center;”></p>
        The creator of Fooocus is not just any developer but the inventor of ControlNet, a crucial subsystem for Stable Diffusion, which revolutionized the world of image generation by allowing artists and designers to fully control the art they create. The creator compares his project to Midjourney in terms of art quality and ease of use. Indeed, the entry barrier to this neural network is very low, and the results are excellent from the first generation. Let’s install, study, and draw conclusions. Let’s go.

        What we will need:

        A computer or laptop with a video card of at least 8GB of video memory.

        About 25GB of free disk space for one mode and 40GB for all three.

        Or a Google account to run in the cloud.

        Fooocus is not yet banned in Google Colab, which means that if you don’t have a suitable computer, you can run the application on Google’s servers completely free of charge.

        Running in Google Colab

        Opening here: https://colab.research.google.com/drive/1ODtrj0vv2Eqxe2NmxI0rwxpgF0JbQVli?usp=sharing

        Wait for the download and installation on the Google Colab server, which may take up to 10 minutes.
        <p style=”text-align: center;”></p>
        You will know that the installation is complete and the program is ready to operate when you see “App started successful.” at the bottom of the console, accompanied by a link like https://some-numbers.gradio.live. You will need to click on this link. The program will open, ready to use.

        Local Installation

        If you have a Windows PC and an NVidia video card, all you need to do is download the archive from this page: https://github.com/lllyasviel/Fooocus#windows by clicking on >>> Click here to download <<< . Unzip the archive into any convenient folder that does not contain Cyrillic characters in its paths.
        <p style=”text-align: center;”></p>
        After you have unpacked the archive, there will be three files in the folder: run.bat, run_anime.bat, and run_realistic.bat. Each file launches the corresponding mode. I will show you the modes visually below, but for now, you can choose whichever you prefer. I will start with the default mode – run.bat.

        For installation on Mac, AMD, Linux, etc., go to the project’s GitHub: https://github.com/lllyasviel/Fooocus page and explore the methods on your own. This does not present any difficulty. I will continue.

        If you’ve done everything correctly, whether locally or on Google, you should now see the Fooocus interface, which looks something like this. Let’s try writing a prompt and see what happens. For me, it will be: “1 woman, short blonde hair, pale skin, medium breasts, pert, glossy skin, in a torn dress, wrists tied to a pole, sweat dripping from her, over-detailed, 8K, absurd, bright colors, depth of field, cinematic lighting, high contrast, golden glow, clear outlines, rich details, humiliating atmosphere, masterpiece BREAK, super detailed, 8K, bright colors, source_realistic bright colors, high contrast, dynamic light, warm glow, clear outlines, rich details, seductive atmosphere“. The first generation will take longer than subsequent ones because additional files are still being downloaded. Here’s what I got.
        <p style=”text-align: center;”></p>
        In my opinion, it’s an excellent result.

        How to Write Queries

        To ensure the neural network understands you, it’s important to learn how to write queries correctly. In Fooocus, we use SDXL models that understand human language very well, and an additional GPT engine autonomously improves your text queries, so you won’t need any special knowledge. Simply describe what you want to see, following this structure: Type of image, object, description of appearance, additional elements, location, effect, style.

        For example: “Scottish folk music is sad, played with bagpipes, flutes, fiddles and harmonicas, a male voice
        (essay on the film Braveheart)”
        <p style=”text-align: center;”></p>
        In my opinion, it’s an excellent result, but don’t be upset if you don’t succeed right away. Writing queries is a skill; practice for just a week and you’ll be doing much better.

        That’s all for today; we will continue tomorrow. Tomorrow’s topic: Launch Modes.

        Sincerely, Imeda Sheriphadze ~Specialist in information technology and software, as well as in neural network design and their application in modern life.

         

    Viewing 0 reply threads
    • You must be logged in to reply to this topic.