Proceeding on Landscape [Realistic/Anime] Series

🌸 Transition to LoRA-Based Releases 🌸

Up until now, I’ve been publishing checkpoint models under the Landscape Realistic / Anime Pro / Max series. It’s been such a joy sharing these with everyone, and I’m truly grateful for all the love and support I’ve received along the way~ 💖🌈

A few months ago, I transitioned all of my systems to beta software versions, which required me to use PyTorch-nightly for any model training or merging. Sadly, this created major inefficiencies — not only did it make it difficult (and sometimes impossible) to train new checkpoints, but merging models also became a headache due to compatibility issues with A1111 / Stable Diffusion WebUI, especially under the MPS backend (MacOS), which has stopped receiving updates.

🥺 As a result, I wasn’t able to continue producing full checkpoints or merge models like before…

But good news! ✨✨✨

I recently got a new laptop with an RTX 5090 Laptop GPU and Intel Ultra 9 processor~ While it’s not exactly the same as training on a full desktop (the laptop 5090 is only half the power of a real 5090), it opens a new door for me:

🌸 I’m now shifting toward LoRA-based subversion releases! 🌸

Instead of training full checkpoints every time, I’ll be releasing LoRA upgrades based on my original models! This allows us to continue building on top of the original checkpoint while keeping things more lightweight and efficient~ 💞💞

🌟 Model Versioning System 🌟

To help keep things organized and easy to understand, I’ve created a universal naming system for both base checkpoints and LoRA submodels:

Checkpoint Naming

• v1.0 — Base

• v2.0 — owo

• v3.0 — uwu

• v4.0 — ;w;

• v5.0 — .w.

LoRA Submodel Naming

• Base 1, Base 2, Base 3 → (1.1, 1.2, 1.3)

• owo 1, owo 2, owo 3 → (2.1, 2.2, 2.3)

• uwu 1, uwu 2, uwu 3 → (3.1, 3.2, 3.3)

• and so on~ 🫧

💫 Introducing a New Model Series: EstellaXL 💫

Furthermore, I’d love to introduce a brand new model line — EstellaXL! 🌟
This is a custom anime model specifically designed for intricate lighting renders, perfect for scenes with cinematic, dramatic, or magical glow ✨💡

EstellaXL is based on IllustriousXL, and is trained separately from the Realistic / Anime series. So if you're into detailed, lighting-heavy compositions — this one’s for you!

As for the Anime and Realistic series: many of the training images I used came from around 2020, which… let’s be honest, aren’t always the highest quality by today's standards 😅 So slowly but surely, I’ll be revisiting them too!

I know this is all a long journey, and right now I’m also preparing for college applications, so I don’t always have time to update everything as fast as I’d like… 😭 But please bear with me 💗 I promise we’ll continue making these models even more beautiful, expressive, and spectacular together~ 🎨🌌

Thank you for all your patience and love~