Daddy. Ink and Masculinity.
Ink and Masculinity. Fusion of Figure, Type, and Form
- 3Jock
- 4Chav
Control Net Series
AI-Enhanced Typographic Design — combining elements of typography, graphic design, abstract expressionism (paint splatter), and photorealistic rendering of a human figure, to create a harmonious composition achieved through the power of AI. These techniques were developed during the creation of the Sacred Geometry series, employing a network of control nets within Stable Diffusion, combined with custom-designed illustrations in Adobe Illustrator.
I have long desired to utilize AI for typography, but like many others have discovered, AI struggles to accurately represent legible type. However, during my work on the Sacred Geometry series—a collection that combines human figures with historically significant geometric forms—I had an enlightening idea. If I could successfully reproduce geometric forms using custom illustrations, why not attempt the same with typography? Here are the results.
In this initial series, the word “Daddy” is integrated with a male figure donning latex and leather pants, set against a background divided by forms with paint splatter. While it may appear as if the type was placed on top of the image after generation, this is not the case. One of the main challenges I faced was presenting the type legibly while ensuring it remains connected to the forms. Ultimately, only a select few instances seamlessly merge the typography with the images. However, in subsequent sets, in which I have refined the technique, you will notice a much stronger integration of type within the composition.
As I previously mentioned, the process involved in creating these artworks is multi-layered. To provide an overview of the workflow, I initially construct a composition within Illustrator, placing the word in a square and selecting an appropriate font and size. Next, I create abstract forms based on the text, resulting in the final image that will serve as the control net. Within Stable Diffusion, I describe the scene using text prompts and employ the Canny control net, which I customize with specific weight, start, and end parameters to influence the composition. Furthermore, I use a different image to guide the color and style, ensuring vibrant colors and the desired expression.
In the remaining two sets, I will showcase this same technique applied to two other words. This represents a significant breakthrough for me as a designer, and I intend to create more of these compositions in the future, marking an important milestone in my creative journey."
10 images on Instagram. 31 images on gymdreams8.com. Some of the images featured on my website have more succesfful, yet I understand that they may not be suitable for all audiences. If you are interested in viewing them, please visit my website. You can find the link in my bio.
Tech: Text prompts with Stable Diffusion: 30 steps, DPM++ SDE Karras, CFG 5, 768x768, Airfuck’s Wild Mix v1, Denoising 0.5. ADetailer 1 mediapipe_face_full, ADetailer 2 hand_yolov8n.pt. CN0 Canny. CN1 Shuffle. Hires 1.5x (1152x1152), 15 steps, 4x_NMKD-Siax_200k. Post: Topaz Gigapixel HQ 2x (2304x2304), Adobe Lightroom post-processing and editing.
Visual Explanation
This infographic shows the technical detail behind how I made these images.
Images
Images
Images
Additional Images
Images
Failed Images
Since this tecnique relies heavily on Stable Diffusion being able to fit the figure into the line art, even though the composition is arguably workable, sometimes the figures may deform in such a way that may not be visually pleasing. I’m ok with one of two body parts moving around because the image is abstract in nature — one might even call them cubist, but when the proportion of the figure is off, they’ll be very noticable. Here’s an example.
Images
