A retro 90s-style anime-aesthetic LoRA trained specifically for the Anima 2B (DiT) architecture.
This model excels at capturing classic hand-drawn cell-shaded textures, clean, sharp line art, and bold, high-contrast hair and clothing highlights reminiscent of late 90s television animation.
Optimized to handle complex props and dynamic perspective shots cleanly without breaking hand or body anatomy.
### Recommended Generation Settings for Anima:
- LoRA Weight: 0.85 - 1.0
- Native Resolution: 768x768 (or bucketed equivalents)
- Classifier Free Guidance (CFG): 3.5 - 5.5
- Base Model: Anima Base v1.0
- Trigger concepts: "retro 90s anime style," "cel-shaded," and "sharp line-art"
Description
Initial release of the Anima-native retro 90s anime-style LoRA.
This model is trained locally on a meticulously captioned 117-image dataset using an optimized low learning rate to fully preserve Anima's unique Qwen-based text understanding. It accurately injects cell-shaded aesthetics, crisp line art, and classic dramatic highlights into modern generation pipelines.
- Retains excellent structural fidelity for handheld mechanical props (e.g., duel disks).
- Highly responsive to background and clothing prompt adaptations.