Seed of Life - Angel
Sacred Geometry Series
- 1Study
- 2Series 1: Mage + Symbology
- 3Series 2: Explanation
- 4Series 3: Angel + Demon, Seed of Life
- 5Technical Details
- 6Series 4: Angel + Demon, Metatron’s Cube
- 7Series 5: Fire Dragon
After posting the Seed of Life - Jock variation, where I took a simpler approach to make the intention of the Sacred Geometry series clear, I thought that I will extend this simpler approach back into the more mystical realm.
The first of this is the Angel seen here. It’s still within the approachable confines in that these angels have wings, but are dressed in more familiar outfits — white jockstraps and white sneakers.
Forms
In terms of how the forms are expressed, I tried to use the opportunity to visually explain how to work with Canny and Line Art visually in creative ways. While you can certainly apply them as part of the layout or do use line art to define the figures themselves, one powerful aspect is that you can turn these lines into actual objects.
In the Jock series, I kept them as mostly lines — laser beams actually. In this Angel set, I transformed these lines into white ropes. The way you would do this is from text prompt. Here what is special about the text prompt are these keywords: white wires, white ribbon, white rope. These are the keywords that will turn the lines into objects.
Environment
As per usual, I like to add dust and particles in my works to give it the texture and depth that adds so much to a scene. So I used light particles and sparks. You can control their influence by using weights. For example: (light particles:1.3) if you want more, or (light particles:0.5) if you want less.
Text weights inside Stable Diffusion are expressed in 0-2, with 1 being the center. While it’s possible to add prompts that are beyond 2, doing so can possibly create so many artifacts that will make the image unusable, so don’t do that. Btw, if you sometimes see your images rendered weirdly, check that you actually added a . in the numbers. That’s a fairly common mistake. Weights and Loras are often the culprits.
Control Net
Two control nets are used: Canny and Line Art. I kept both at weight 1 (100%), but the important detail is that I start at 0.5 and end at 1. If you start the CNs at 0, the lines will restrict the character to be formed properly, because Stable Diffusion will usually then try to insert the figures only in the empty areas where there are no lines. By starting the lines midway through the steps, you can ensure that the forms be formed on top of the figure. Ending at 1 ensures that the lines will be fully visible at the end of the image generation.
29 images total on gymdreams8.com. IG sets include less images, because of reasons.
Tech: Text prompts with Stable Diffusion: 30 steps, DPM++ SDE Karras, CFG 5, 512x512, Virile Fusion v2, Denoising 0.5. ADetailer with prompts face_yolov8n.pt. CN0 Canny (Weight 1, Starting 0.5, Ending 1, Res 1024). CN1 Line Art Standard (Weight 1, Starting 0.5, Ending 1, Res 1024). Hires 1.5x (768x768), 15 steps, 4x_NMKD-Siax_200k. Post: Topaz Gigapixel HQ 4x (3072x3072), Adobe Lightroom color correction, Photoshop AI Generative Layers (limb fixes).
Images
Set 1
Set 2
Set 3
Image Index
I order my images as sets of ratings high to low according to my own personal taste. For series that have 3 sets, it could be hard to visually distinguish them, as each set is rating 5 to 1. In the image index, all the images are ranked from 5 to 1 collectively. It’s also the better way for someone to let me know which image they are interested in, so that’s why they are ordered this way.
