Welcome to AIcAI - AnimeHaven✨
The first CivitAI model to use ILLUSTRIOUS V1.1
( The following is written by my hands and a mix of AI )
Feel free to send feedback !
🎉 New Version Available: AIcAI - AnimeHaven V20! 🎉
Building on the foundation of previous versions, V20 offers noticeable improvements:
Enhanced Sharpness & Detail: Achieve even crisper lines and finer details in your creations.
More Vibrant & Deeper Colors: Experience richer, more saturated color palettes that make images pop.
Improved Lighting & Contrast: Generate images with more dynamic lighting and striking contrast, including excellent monochrome capabilities.
V20 aims to provide greater polish and refinement for your anime-style illustrations.
Recommended Settings:
Here are the settings I've found work well during testing. For optimal results, please try the following:
Dimensions: 768x768 all the way to 1536x1536
Steps: 25 to 50
CFG Scale: 3.5 to 6.0 (for best results)
Sampler: Euler_a, DPMPP2M
Scheduler: Karras, sgm_uniform
Hires. Fix: Strongly recommended.
Hires steps: Around 10-20 often works well.
Denoising strength: Start between 0.35 - 0.55. Lower values preserve the original composition more, higher values add more detail/change. Adjust based on your results.
Notes:
As this is a merge, results might lean towards specific aesthetics. Experimentation with prompts is encouraged!
Releases notes from previous versions.
🎉 New Version Available: AIcAI - AnimeHaven V11! 🎉
A new version, AIcAI - AnimeHaven V11, is now available !
Compared to V10, V11 offers:
✅ Improved Prompt Control: Better adherence of prompts and sentences.
🎨 Enhanced Colors: More vibrant colors.
Check it out for the latest improvements!
Hello everyone! I'm excited to share my very first model with the community, AIcAI - AnimeHaven.
This is a checkpoint merge aimed at generating high-quality, vibrant anime-style illustrations. I've focused on achieving good color, detail, and an overall polished look, particularly for character-centric scenes. ILLUSTRIOUS 1.1 made it possible so that my model can generate accurate images without hallucinating with dimensions up to 1536x1536 (even more in some cases).