Overview

Lewdiculous/lwd-Mirau-7b-RP-Merged is a 7.6 billion parameter model, a merged LoRA adaptation of the Qwen/Qwen2.5-7B-Instruct base model, developed by mouseEliauk. It is specifically designed for first-person narrative generation, transforming simple user actions into vivid storytelling. The model employs a unique "story flow chain of thought" training method to ensure continuous and coherent narrative progression, making it suitable for immersive text-based experiences.

Key Capabilities

• Action-to-Narrative Render: Converts user inputs (basic actions, exploration, inner monologue, observation) into detailed narratives.
• Immersive Storytelling: Generates environmental descriptions, psychological activities, and plot developments.
• Coherent Narrative Flow: Utilizes a "story flow chain of thought" method to link user inputs with previous context.
• Multilingual Support: Capable of processing and generating text in numerous languages including Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, and Arabic.

Important Notes & Limitations

• First-Person Perspective Only: The model is exclusively designed for first-person narrative.
• Not for Dialogue: It is not suitable for dialogue interactions, as outputs may be unstable.
• Single-Character Focus: Best used for single-character narrative experiences.

Good For

• Text-based adventure games.
• Mystery story generation.
• Exploring imaginative scenarios through narrative.
• Applications requiring detailed, continuous first-person storytelling from user actions.