ReadyArt/Serenity-31B-v1.1

Hugging Face
VISIONConcurrent Unit Cost:2Model Size:31BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 10, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

ReadyArt/Serenity-31B-v1.1 is a 31 billion parameter language model based on the GEMMA-4 architecture, fine-tuned using LoRA for enhanced depth and balanced tone in adult-oriented roleplay. It harmonizes male-to-female and female-to-male perspectives from its Melody and Darkside lineages, creating nuanced character interactions. Optimized for immersive, mature-themed dialogue, it leverages advanced Character and Emotional Engines for consistent personality and realistic emotional responses. The model is specifically designed for personal, non-commercial use in 18+ content scenarios.

Loading preview...

Serenity-31B-v1.1: A Hybrid Model for Immersive Adult Roleplay

Serenity-31B-v1.1 is a 31 billion parameter language model developed by ReadyArt, built upon the GEMMA-4 architecture. This model is uniquely designed for adult-oriented roleplay, harmonizing the expressive depth of "Melody" (male → female perspective) and the immersive presence of "Darkside" (female → male perspective) into a single, balanced entity. This hybrid approach aims to provide a refined personality and natural narrative flow, though it may occasionally exhibit perspective confusion due to its merged lineages.

Key Capabilities & Features

  • Dual Heritage: Combines opposing perspective lineages for a balanced tone in interactions.
  • Synthetic Life Engine: Utilizes a proprietary Character Engine for consistent personality traits and speech patterns, and an Emotional Engine to inject dynamic, realistic emotional states.
  • Quality Refinement: Automated systems detect and rewrite repetitive phrases, ensuring fresh and high-quality dialogue.
  • Dialogue Integrity: Advanced quote normalization prevents formatting errors and maintains immersion.
  • LoRA Fine-tuning: Efficiently adapted using LoRA (Rank 80, 2 epochs) to imprint new behavioral patterns while preserving the base model's capabilities.
  • Specialized Dataset: Trained on a strictly 18+ adult-oriented roleplay dataset focusing on mature themes and uncensored dialogue.

Good For

  • Immersive Adult Roleplay: Excels at generating nuanced, emotionally resonant, and uncensored dialogue for mature themes.
  • Character Consistency: Maintains consistent personality and behavioral logic across extended interactions.
  • Dynamic Emotional Responses: Creates realistic and dynamic emotional states within character interactions.

Important Considerations

  • Content Advisory: This model is intended for mature audiences (18+) and can generate explicit content.
  • Perspective Note: Due to its hybrid nature, occasional perspective confusion may occur.
  • Usage: Licensed under Apache 2.0 for personal, non-commercial use only. Users bear full responsibility for generated outputs.

Finetunes
2 models