This model will help you draw vivid mosaic style animals.
#### Model Overview
- Type: LoRA (Style)
- Base Model: SDXL 1.0
- Version: 3.0 (Gen 3 - Refined with expanded dataset for better dynamics and color variety)
- Trigger Words: zmosaic (primary trigger; use at the start of prompts for strong style activation, e.g., "zmosaic psychedelic panther leaping")
- Secondary Trigger Words: simple_background, anthro, mosaic, hexagon, circuit, portrait, full_body,
- Recommended Strength: 0.7-1.0 (0.8 for balanced mosaic grids and glows; lower for subtle integration with other styles)
- Training Details: Trained on 162 original 1024x1024 PNGs (curated from original and iterative gens), using Civitai's on-site trainer, rank 48, alpha 24,
#### Description
Cyber-Mosaic Animals is a specialized LoRA that transforms any animal subject into a mesmerizing psychedelic mosaic masterpiece. Inspired by glowing geometric grids, vibrant hexagonal patterns, and surreal color gradients, this style evokes a cyber-Aztec fusion—think abstract animal forms with luminous outlines on dark, cosmic backgrounds. Perfect for creating digital art, NFTs, or trippy visuals, it excels in rendering dynamic poses (leaps, sprints, flights) with flowing distortions in the mosaic tiles.
Built iteratively across three generations:
- V1: Initial capture from 82 core images, focusing on basic grid structures.
- V2: Added refinements for glow intensity and animal variety.
- V3: Expanded with 80+ generated images emphasizing movement and environments, resulting in superior flexibility and fidelity.
This LoRA plays well with realistic or anime checkpoints—merge via tools like ZipLoRA for hybrid styles. No additional inference latency, and it's memory-efficient for consumer GPUs. If outputs feel muted, boost with noise offset 0.1 or CFG 6-8. Tested on A1111 and ComfyUI; clip skip 2 recommended for SDXL.
#### Usage Instructions
- Prompt Structure: Start with "zmosaic [animal/subject], [action/description], psychedelic grids, glowing hexagons, vibrant colors, dark background".
- Negative Prompts: blurry, realistic, lowres, overexposed, text, undetailed (to maintain abstract mosaic vibe).
- Settings: Resolution 1024x1024, steps 30-50, sampler DPM++ 2M Karras, CFG 6-7.
- Merging Tips: For advanced blends, use weights like <lora:zmosaic_v3:0.8> + another style LoRA at 0.4. Code snippet for Diffusers:
```python
from diffusers import StableDiffusionXLPipeline
import torch
pipe = StableDiffusionXLPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.bfloat16)
pipe.to("cuda")
pipe.load_lora_weights("zmosaic_v3.safetensors")
image = pipe("zmosaic psychedelic cat leaping, glowing orange grids, dynamic motion trails", clip_skip=2, num_inference_steps=40, guidance_scale=6.5).images[0]
```
#### Example Gallery
1. Orange Mosaic Cat (Walking Pose)
Prompt: "zmosaic orange cat strolling dynamically, hexagonal grids with glowing outlines, blue eyes, dark textured floor background"
Settings: Strength 0.8, seed 42, noise offset 0.05.
2. Blue Mosaic Bear (Sitting Profile)
Prompt: "zmosaic blue bear sitting alert, tile mosaic patterns, white eyes, neutral gray background" Settings: Strength 0.9, seed 123, CFG 7.
3. Red Mosaic Deer (Head-On View)
Prompt: "zmosaic red deer gazing forward, intricate hexagonal grids, black eyes, deep red background"
Settings: Strength 0.85, seed 456, resharpen -0.5 for soft glows.
#### Tags
LoRA, Style, Psychedelic, Mosaic, Animals, Geometric, Glowing, Abstract Art, SDXL, Digital Art, Cyberpunk, Surreal
If you enjoy Zmosaic be sure to give this model a "like" and check out similar styles on my creator page or share your creations below! Feedback welcome for future iterations.