JAX Randomizer Project, Hyper-SD, 2024-06-06 Vol. 020
The Creation of 5000 Unique Images: A Journey with Fooocus and Wildcards
The Story Behind the Project
Recently, I embarked on an exciting journey to create a series of unique and captivating images using the power of AI, specifically Fooocus. The adventure began with a collaborative effort between myself and ChatGPT 4.0, aiming to explore the depths of digital art through the use of meticulously crafted prompts and an extensive set of wildcards.
The Wildcards
To achieve the desired variety and depth in the images, we created a comprehensive set of 39 wildcards, each representing different aspects of visual and contextual elements. Here are the wildcards we used:
Aesthetics: colours, general aesthetics
Appearance: adjectives, appearance, size
Entities: entities primary, entities secondary, mood, nationality, occupations
Materials: building materials, common metals, decorative stones, fabrics, fibres, gems, metals, packaging materials, technical fabrics
Modifiers: camera, lightning, material, misc, photography, pop culture, punkism, repository, time period, views
Places: home, locations
Style: art movement, genre, platform, styles
Time: decade, seasons, timeofday
Misc: vehicles
Each wildcard served as a placeholder within the prompts, allowing Fooocus to randomly select and incorporate elements from the predefined categories, thereby ensuring a unique output every time.
The Prompts
The core of our image generation process revolved around creating detailed and varied prompts. Here is an example of a prompt used:
Example Prompt:
jax_RND--entities---entities_primary in jax_RND--places---locations with jax_RND--appearance---adjectives, jax_RND--appearance---size, and jax_RND--modifiers---lightning, jax_RND--materials---fabrics, jax_RND--style---art_movement
This prompt includes primary entities, locations, appearance descriptors, materials, and a fixed art movement to ensure a visually captivating and contextually rich image.
The Result
Using Fooocus, I managed to generate nearly 5000 unique images in just 24 hours. Each image is a testament to the versatility and depth that can be achieved through a well-constructed set of wildcards and thoughtfully crafted prompts. The level of detail and variety in the images was astounding, and the creative possibilities felt endless.
This process allows me to test various combinations of checkpoints and LoRAs intensively. It also enables me to develop a specific style or scene and build a series of images based on that foundation. Each image is random, but I control what is depicted through the content of the wildcards.
Technical Setup
My setup played a crucial role in the efficiency and quality of the image generation process. Here's a brief overview:
OS: Garuda Linux BirdOfPrey Soaring x86_64
Host: X670E AORUS PRO X (-CF)
Kernel: Linux 6.9.1-zen1-1-zen
CPU: AMD Ryzen 9 7900X3D
GPU: AMD Radeon RX 7800 XT
Memory: 48 GiB
DE: KDE Plasma 6.0.4
WM: KWin (Wayland)
The settings for most images remained consistent, utilizing the following LoRAs and checkpoints over 90% of the time:
LoRAs: Detail Tweaker XL v1.0, SDXL FaeTastic Details v1.0, ral-ledlights-sdxl, ral-dstgrtptrn, Perfect Eyes XL v1.0, sdxl_hyper_sd_4step_lora
Checkpoint: albedobond/AlbedoBase v2.1
I focused on the composition of the images and enjoyed the results, occasionally influencing the structure using PyraCanny, ImagePrompt, and FaceSwap.
Join Me on CivitAI
https://civitai.com/user/JohnAndrewsX
I'm relatively new to the CivitAI community and would love to hear your thoughts on these images. More pictures from this series can be found on my profile. Your feedback and follows would be greatly appreciated, as it would encourage me to continue exploring and creating more digital art.
Join Me on X
Additionally, you can join me on X where I share updates, behind-the-scenes insights, and more artwork from this and other projects. Follow me for regular content and engage with a community passionate about AI and digital art. Your support on X would mean a lot and help me grow as a creator.
Thank you for being a part of this journey, and I hope you enjoy the images as much as I enjoyed creating them!