This LoRA is designed to capture the "DicPic Deliverer" character with a focus on rigid structural integrity, matte resin-like skin texture, and precise brand placement.

Recommended Usage

• Base Model: Pony Diffusion V6 XL (This LoRA is specifically tuned for this architecture).

• LoRA Weight: Recommended 0.8–1.0.

• Trigger Word: dicpic_deliverer

• Prompting Style: Use the trigger word followed by character descriptors (e.g., short red jacket, mailman cap, resin skin) combined with standard Pony-style tags for poses and actions.

Recommended Generation Settings (Ksampler)

To maintain the intended aesthetic and avoid "spaghetti" limb distortion:

• Sampler: Euler a or DPM++ 2M Karras

• Steps: 25–35

• CFG Scale: 5.0–7.0 (Pony models generally perform best at lower CFG ranges).

• Resolution: 1024x1024 (Native SDXL resolution is optimal).

Tagging Guide

The LoRA is trained with the following core tags. Use these for best results:

dicpic_deliverer, short red jacket, mailman cap, resin skin, grey skin, hand truck logo on cap, hand truck logo on belt, dicpic logo on jacket, single white pouch on each side of waist, black reflective visor, helmet, boots, belt, briefs, uniform

Negative Prompting Guidelines

To ensure character consistency, it is highly recommended to use the following negative prompts:

removing helmet, showing face under the helmet, appearing as a mascot, inner suit, appearing as robot, missing cap, spaghetti limb flailing, anatomical blurring

One final tip: Since you are using Pony, remember that the "Score Tags" (e.g., score_9, score_8_up, score_7_up) are often required by the base model to ensure high-quality output. Make sure to include them in your prompts if your specific workflow requires them.