Summary
This Illustrious-based model merges multiple checkpoints and specializes in anime style generations. Optimized for SDXL, it delivers the best results with Euler sampling and supports LoRAs for enhanced customization.
Ideal for: Anime art, character design, and vibrant illustrations.
Overview
Key Features:
✅ SDXL-optimized for anime aesthetics
✅ WebUI & ComfyUI compatible
✅ LoRA-friendly (tested with my SDXL-specific LoRAs)
Recommended Parameters:
CFG Scale: 3 – 5
Steps: 20 – 35
Sampler: Euler (best results). Other samplers may not perform as intended, but experimentation is encouraged.
Optimal Resolutions:
1024 × 1024 (1:1)
1254 × 836 (3:2)
1365 × 768 (~16:9)
License & Restrictions:
❌ Non-Commercial Only (No selling generated images)
⚠️ Ethical Use Required (No harmful/illegal content)
Merge Details:
Model Base: SDXL (Stable Diffusion XL)
Merged with a weight of 0.45 on:
VividIllustrious_V1.5bf16
License & Usage:
❌ Non-Commercial Only (No selling generated images)
Commercial Use Prohibited: You may not sell images generated from this checkpoint.
⚠️ Ethical Use Required (No harmful/illegal content)
Legal & Ethical Note: Do not use this model to create harmful, illegal, or hateful content. Users assume full responsibility for any consequences arising from its usage.
Checkpoint License: Illustrious
Description
FAQ
Comments (7)
is bf16 version better somehow?
bf16 it’s a good option because it can handle a wider range of numbers, even if it’s not as precise with the smaller details. This makes bf16 useful when working with very large numbers or when it’s more important to keep it steady across many sizes of numbers rather than being exact. In short, BF16 covers a larger range with less detail, making it better for tasks that need to handle a wide range of numbers or maintain stability, while FP16 is better at showing more detail with smaller numbers but operates within a smaller range.
Happy to see this getting an update!
Hope we can see the other checkpoints getting some love too.
Currently working on improving my other check points as well, still preparing the data set
very seductive model, not very prompt accurate
Hi, what Parameters are you using : CFG Scale, Steps , Sampler?
Are you using WebUI & ComfyUI or other ?
Depends on your chosen prompt, certain concepts or poses may need Lora's
@Rustyvivid 4.5/32/dpm2m/comf/lora(+)
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.