import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("krea/Krea-2-Turbo", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("WarmBloodAban/Krea2-Anime-Style-Collection")
prompt = "-"
image = pipe(prompt).images[0]Krea2-Anime-Style-Collection
- Prompt
- -
Model description
license: openrail tags: - text-to-image - stable-diffusion - lora - anime - comfyui - krea2
Krea2 Anime Style LoRA Collection
> ### 📢 Continuous Updates & Iterations > This repository is actively maintained. New model variants, expanded datasets, and optimized checkpoints will be continuously updated and iterated over time to ensure peak generation performance.
A curated collection of Anime Style LoRA models engineered for high-end illustrative anime art, cinematic global illumination, and dynamic diagonal composition layouts.
📦 Model List & Preview Gallery
| LoRA Name | Recommended Weight | Trigger Words | Visual Effect Preview |
|---|---|---|---|
| `Krea2_DreamWeaveHanfu` | `1` | `trigger words here` |  |
| `Krea2_WaterColor_Iwano_Esthetic` | `1` | `trigger words here` |  |
| `Krea2_AnimeColor_NeonCrystalArt` | `1` | `trigger words here` |  |
| `Krea2_MangaTint_BistroFresh` | `1` | `trigger words here` |  |
| `Krea2_MangaTint_KFashionDesign` | `1` | `trigger words here` |  |
| `Krea2_MangaTint_HeartbeatPrelude` | `1` | `trigger words here` |  |
🚀 Core Features & Environment
- ComfyUI Workflow Native: Fully optimized for the ComfyUI architecture. Extensively tested for line smoothness, clean color gradients, and excellent compatibility with mainstream base models—perfect for embedding into complex automation workflows.
- Krea2 Turbo Parameter Alignment: All LoRAs in this repository are fine-tuned to match the exact recommended inference weights and prompt responsiveness of the Krea2 Turbo engine.
- Premium Visual Aesthetics: Inherits highly recognizable visual anchors, including sharp cross-shaped starburst flares, iridescent lens glints, low-key charcoal bases, and intricate macro details.
⚙️ Recommended Inference Parameters
To perfectly replicate the crisp, high-fashion quality of Krea2 Turbo within ComfyUI, apply the following node configurations:
- LoRA Weight: `0.7 - 1.0` (Default `0.8` is highly recommended for optimal balance)
- Sampler: `Euler` / `DPM++ 2M Karras`
- Steps: `20 - 30` (Optimized for Krea2 lightning-fast inference)
- CFG Scale: `3.5 - 5.0` (Maintains high contrast and crystal-clear generation)
🎨 Krea2 Golden Prompt Blueprint
Add these high-density, pure positive visual phrases to your ComfyUI `CLIP Text Encode` node to maximize the model's stylistic output:
```text [Subject/Character Description], high-angle 3:4 illustration, dynamic diagonal composition, dark low-key charcoal base, brilliant cross-shaped starburst flares, iridescent lens glints, diamond reflections, sharp facial focus, premium high-fashion magazine aesthetic.
Download model
Download them in the Files & versions tab.
🤝 Community & Business
- YouTube Tutorials: https://www.youtube.com/@AIGC-Singularity
- Bilibili Channel: https://space.bilibili.com/49766729
- QQ Group 1: 1058747239 (Leave a message to join)
- QQ Group 2: 1072010342 (Leave a message to join)
- Business Inquiries: aigctyd (WeChat)
- Email： a592991299@gmail.com
- Downloads last month
- 217