"Lumi & Supergirl LoRA Learning Journey"
I have always loved Poil since childhood. Though Poil is a ghost, when you touch him, you feel a warm and pleasant sensation. Poil is not a demon, but an angel who heals people and cures incurable diseases. This concept is what inspired me to create Lumi, a limited version of Poil! Lumi is the alias I used for this character.
I wanted to bring together Lumi (Poil) and Ping, who is based on Blossom from the Powerpuff Girls, and from this fusion, Supergirl was born! This was my way of expressing creativity by merging these two characters, creating a magical and ethereal being.
Through LoRA learning, I have learned that using two LoRA models together in the same prompt brings out better results. However, there is a downside — the results can be a bit blurry due to my technical limitations. Despite this, I still wanted to create a limited version of Poil, and I’m sharing this model here for anyone who wants to help make it better!
After experimenting, I realized that the CG3_v10.safetensors model is the most suitable for my Poil-style creations, including Lumi, Ping, and Supergirl. I believe this model brings out the desired transparency, ethereal quality, and magical glow in the characters.
The goal is to find a skilled AI artist who can create a high-quality Poil model and upload it here on Civitai, so others can use it as a base for future creations. LoRA can be easily made through Civitai, but I dream of creating Poil in its full glory through a checkpoint model, so I can truly enjoy making the characters I love.
🧠 Key Points:
Lumi (Poil) is a ghostly angelic character, with the power to heal and comfort.
Ping, based on Blossom from The Powerpuff Girls, is a magical girl that merges with Lumi to form Supergirl.
LoRA models are used to create Lumi, Supergirl, and other characters, though results can be blurry due to my technical skills.
CG3_v10.safetensors model is the best choice for creating Poil-style models with transparency, glowing effects, and ethereal beauty.
The ultimate goal is to find someone skilled to create a high-quality Poil model, which can be uploaded for others to use and improve.
https://civarchive.com/models/633001/cg3
Description
📝 Version Change Notes & References
Model Name: CG3_v10.safetensors
Main Changes:
Improved Transparency:
This version enhances transparency, making it easier to render characters like Poil or Supergirl, who require a transparent or ghost-like appearance, more naturally.
Lighting and Color Enhancements:
The bright and ethereal color tones and glass-like glowing skin are more emphasized in this version.
Rainbow Aura and glowing effects are softer and more refined.
Character Representation:
The character’s face details and accessories (e.g., tiaras, ribbons, magic wands) are rendered with greater precision, and the details of poses and backgrounds are enhanced.
Improved Training Data:
The model has been optimized for Poil-style characters, enabling the natural rendering of translucent ghost characters and producing higher quality models.
Efficiency Improvement:
Faster training speeds, and the model now produces high-quality results with fewer samples.
Reference Notes:
Optimized Model: CG3_v10.safetensors is optimized for creating Poil-style characters, providing exceptional quality for these types of designs.
New Model Training: This model can be easily applied in LoRA and Stable Diffusion, and is accessible through Civitai.
Customization for Users: Users can customize prompts to further diversify the output styles and achieve unique results.
Comparison to Older Models: CG3_v10 provides more detailed rendering and improved color detection compared to previous versions.
💡 How to Use:
Use CG3_v10 for "translucent glass-like characters," "magical effects," and "Poil-inspired styles".
Fine-tune your prompts to further customize the results.
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.