Today's progress 🥳🎊 (Patreon)

Today's progress looks really promising for the future!

Today, I worked on creating several pose and concept models that helped me understand my current limitations better.

The models I worked on were:

• U.A Cheerleader uniform

• Lying Legs behind head

• Buttjob under panties

• Lipstick Mark On penis

• Floating Onomatopoeia

At first glance, these might seem like simple concepts that wouldn’t make a huge difference in my images or content. But in reality, each of these models has allowed me to create subtle and highly specific details that serve as a stepping stone for a complete shift in quality and variety.

Here’s a breakdown of what I achieved with each model:

U.A Cheerleader uniform – A complete improvement in clothing details

This model helped significantly with learning how to generate complex clothing details, including patterns, logos, and text. Now, outfits with intricate designs—like those from Genshin Impact characters—can be reproduced with exceptional accuracy.

This doesn’t just improve the clothing; it also enhances the character’s overall appearance, making them look much more like themselves. The improvements extend to the face, body, and even the character’s overall style.

Lying Legs behind head – A major upgrade in generating complex poses

Previously, creating complex poses was quite limited by how close the arms and legs were to the torso, as well as the number of characters in the image. (This is a simplified explanation, but it’s the best way to put it without getting too technical.)

Now, with this improvement in pose learning, subtle deformations in regular poses will no longer be an issue. This also opens up the possibility of generating much more complex and dynamic poses—ones that, until now, were only possible in non-AI-generated artwork. Unusual angles, detailed close-ups, and an incredible level of precision are now a reality.

Buttjob under panties & Lipstick Mark On penis – Fine-tuning subtle details

Until now, I had never been able to create concept models that modify very specific parts of an image in a subtle way. Previous models had to be oversaturated with training data to force the AI to learn a pose or concept, which often resulted in ruined images or models that didn’t actually learn anything useful.

But with these new models, I’ve managed to introduce subtle modifications with such precision that I was even able to generate transparency in clothing. This might not seem like a big deal at first, but it’s a huge breakthrough. Being able to create such incredibly specific changes at this level of quality means that the current limitations in model training are much further away than I previously thought.

Floating Onomatopoeia – INSANE!

The last three images in this post—the ones of Himiko Toga—speak for themselves. The text you see in those images, while highly stylized and not perfect, was generated along with the image itself. It wasn’t added later with any external software.

In other words, I’ve successfully created a model that can generate readable, stylized, and properly placed sound effects in an AI model that normally can’t produce text.

OH YES!!! having managed to create a model that can reproduce something as abstract, complex, and specific as text means that now there is no pose, concept, or modification that I can't achieve.

What’s coming this year is going to be absolutely wild! 🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥