🧪 Model Info
This model was trained on 140+ images using the Niji Default mode.
It produces results that are bright, vivid, and elegant, with good generalization and flexible weight control.
🔧 Recommended setup: Use with negative embeddings like aid291, badhandv4, etc., for cleaner outputs.
⚠️ This is an experimental version — the dataset is relatively small, but the results are still impressive, especially for duo character compositions.