This LoRA is used to generate Midjourney-style images, leaning toward a realistic film style. Recommended strength: 0.5-0.7. CFG: 3.5, 25 steps. trigger word: mdv7.
Due to nearly unsolvable edge artifact issues inherent in SD3.5L LoRA training, follow these guidelines for parameter settings:
Keep LoRA strength below 0.7.
Prefer square or vertical images (horizontal images, especially ultra-wide ones, will inevitably show edge artifacts).
Slightly reduce image dimensions to multiples of 64 to mildly alleviate horizontal image artifacts (cannot be fully eliminated).