README.md

---
tags:
- text-to-image
- lora
- diffusers
- template:diffusion-lora
- flux
- nft
- pfp
- art
- cartoon
widget:
- text: 'YellowGang, A front-facing view of the character from image_0.png wearing a back-turned black snapback cap, a red basketball jersey with "1" and "YellowGang" logo, gold chains, and holding a small basketball, solid blue background.'
  output:
    url: >-
      https://cdn-uploads.huggingface.co/production/uploads/no-id/placeholder-yellowgang-default.png
- text: 'YellowGang, A character from image_0.png wearing a futuristic cyborg motorcycle helmet with a glowing red LED visor, an astronaut suit with the "YellowGang" patch, and a starry cosmic background.'
  output:
    url: >-
      https://cdn-uploads.huggingface.co/production/uploads/no-id/placeholder-yellowgang-cyborg.png
- text: 'YellowGang, The yellow character from image_0.png clean-shaven (bald), wearing a classic black and white zebra bandana, sleek black aviator sunglasses, and a pink Japanese Kimono with the "YellowGang" logo.'
  output:
    url: >-
      https://cdn-uploads.huggingface.co/production/uploads/no-id/placeholder-yellowgang-kimono.png
- text: 'YellowGang, A close-up portrait of the character from image_0.png with large over-ear DJ headphones, angry red eyes, a spiked leather choker collar, and a blue streetwear hoodie with the "YellowGang" logo, dark city background.'
  output:
    url: >-
      https://cdn-uploads.huggingface.co/production/uploads/no-id/placeholder-yellowgang-punk.png
- text: 'YellowGang, A character from image_0.png with neon green spike hair, "X" dead-style eyes, a tattered military uniform, and a zombie-groan mouth expression.'
  output:
    url: >-
      https://cdn-uploads.huggingface.co/production/uploads/no-id/placeholder-yellowgang-zombie.png
base_model: black-forest-labs/FLUX.1-dev
instance_prompt: YellowGang
license: creativeml-openrail-m
---

![yellowgang_banner.png](https://cdn-uploads.huggingface.co/production/uploads/no-id/placeholder-yellowgang-banner.png)

<Gallery />

# Model description for YellowGang-Generator-LoRA

This LoRA model is a specialized generator for the "YellowGang" collection, based on the signature yellow-skinned character style with bold cartoon outlines, as seen in `image_0.png`. It is fine-tuned to preserve the character's facial features and proportions while allowing for high-fidelity generation of diverse traits, including various headwear (bandanas, crowns, helmets), eyewear (cyborg, classic, sleepy), outfits (hoodies, jerseys, armor, suits), and neck accessories (chains, chokers, scarfs). The model ensures the "YellowGang" logo is consistently applied to all outfits.

Image Processing Parameters (Used for Training)

| Parameter                 | Value    | Parameter                 | Value  |
|---------------------------|----------|---------------------------|--------|
| LR Scheduler              | constant | Noise Offset              | 0.03   |
| Optimizer                 | AdamW    | Multires Noise Discount   | 0.1    |
| Network Dim               | 128      | Multires Noise Iterations | 10     |
| Network Alpha             | 64       | Repeat & Steps           | 25 & 3500 |
| Epoch                     | 15       | Save Every N Epochs       | 2      |

    Labeling: florence2-en(natural language & English) with specific trait tagging.
    
    Total Images Used for Training : 20 (Base character + variation set)

## Best Dimensions & Inference

| **Dimensions** | **Aspect Ratio** | **Recommendation** |
|-----------------|------------------|---------------------------|
| 1024 x 1024     | 1:1 (Square PFP) | Best (Default for PFP)   |
| 1280 x 832      | 3:2              | Recommended               |

### Inference Range

- **Recommended Inference Steps:** 30–35

## Setting Up

To use this LoRA in your Diffusers pipeline, ensure you use the `YellowGang` instance prompt.

```python
import torch
from diffusers import DiffusionPipeline

# Base model on which the LoRA was trained
base_model = "black-forest-labs/FLUX.1-dev"
pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16)

# The location of the fine-tuned LoRA
lora_repo = "your_username/YellowGang-Generator-LoRA"
pipe.load_lora_weights(lora_repo)

# Trigger word to activate the specific character identity
trigger_word = "YellowGang"  

device = torch.device("cuda")
pipe.to(device)