Unlike standard weighted merging, this method uses SDXL Base 1.0 as a neutral anchor to isolate and extract specific aesthetic deltas from specialized models. This minimizes model-induced artifacts while maximizing texture fidelity and structural integrity.

2. Technical Methodology

This model was constructed using a custom differential summation script to aggregate specific aesthetic components from five distinct models.

Merge Composition:

Anchor Model: SDXL Base 1.0
Differential Weights:
- astroreal_v10.safetensors: 35% (Structural Realism)
- miyabismmixPony_v60.safetensors: 25% (Stylized Balance)
- waiSimpleuseFor_v10.safetensors: 25% (Prompt Adherence)
- wfa_v10.safetensors: 10% (Fine Detail Enhancement)
- evalisenniaRealisticEastAsian_v40.safetensors: 5% (Subtle Skin Texture)

Summation Logic:

The final weights were calculated by isolating the delta of each model from the base anchor: Result = SDXL_Base + Σ((Model_n - SDXL_Base) * Weight_n)

By adding these calculated deltas back to the SDXL Base 1.0, the model inherits the superior composition flexibility of Pony while imbuing it with SAKURA's signature matte textures, transparent skin tones, and physically accurate lighting control.

3. Recommended Settings

Base Model: Pony Diffusion V6 XL
Sampler: Euler a / DPM++ 2M Karras
Steps: 25 - 35
CFG Scale: 5.0 - 7.5
Negative Prompt: (score_4, score_5, score_6, censored, 3d, render, cartoon, anime)

4. About FjikoAI

FjikoAI focuses on the fusion of "Logical Technical Control" and "Emotional Aesthetics." We utilize PBR (Physically Based Rendering) principles to optimize lighting and texture in AI-generated imagery.

🌟 Join the Community

If you enjoy this model, please consider sharing your work and leaving a review on Civitai!

Civitai

FjikoAI/SAKURA_Pony_v0.3

SAKURA Pony v0.3

1. Concept: "Pure Aesthetic Extraction"