IggyLux/MN-VelvetCafe-RP-12B-V2

TEXT GENERATIONConcurrency Cost:1Model Size:12BQuant:FP8Ctx Length:32kPublished:Mar 10, 2026License:cc-by-4.0Architecture:Transformer0.0K Open Weights Cold

MN-VelvetCafe-RP-12B-V2 by IggyLux is a 12 billion parameter language model merged using the SLERP method, specifically designed for immersive, long-term multi-character roleplay. It excels at consistent scene, position, and clothing tracking, maintaining balanced emotions, and reliably handling author's notes and system prompts. This model is optimized for visually detailed and consistent roleplay experiences without degradation over long conversations, making it ideal for interactive storytelling applications.

Loading preview...

MN-VelvetCafe-RP-12B-V2: Enhanced Roleplay Model

MN-VelvetCafe-RP-12B-V2 is a 12 billion parameter language model developed by IggyLux, created through a SLERP merge of Dans-PersonalityEngine-V1.3.0-12b and Neona-12B. This version addresses and fixes issues from previous releases, including incorrect sampler settings and tokenizer mismatches, resulting in significantly improved consistency and reduced degradation in long roleplay sessions.

Key Capabilities

  • Exceptional Consistency: Maintains stable formatting and quality over extended roleplay, preventing common issues like repetitive patterns or forced prose.
  • Immersive Scene Tracking: Excels at tracking scene details, character positions, and clothing, enhancing the overall immersion in long-term roleplay scenarios.
  • Balanced Emotional Responses: Generates narrative-appropriate emotions without random aggression or refusals, ensuring stable character behavior.
  • Reliable Prompt Obedience: Demonstrates strong adherence to author's notes and system prompts, allowing for precise control over the narrative.
  • Multi-Character RP Focus: Optimized for multi-character group chat roleplay, effectively managing interactions between two or more characters.

Good for

  • Developers and users seeking a robust model for long-form, multi-character roleplay.
  • Applications requiring high consistency in character personality and scene details over extended interactions.
  • Scenarios where precise control via system prompts and author's notes is crucial.
  • Users looking for a model that provides visually detailed and emotionally stable narrative generation.

Finetunes
3 models