SlerpE/CardProjector-27B-v4

VISIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kPublished:Apr 10, 2025License:gemmaArchitecture:Transformer0.0K Cold

SlerpE/CardProjector-27B-v4 is a 27 billion parameter language model developed by WoonaAI, fine-tuned from Google's Gemma 3 architecture. This model specializes in generating detailed character cards for SillyTavern and general character creation, with a strong emphasis on personality development using MBTI and Enneagram profiles. It features reduced positive bias and supports vision capabilities, making it highly sensitive and responsive to user prompts for realistic character generation.

Loading preview...

CardProjector-v4: Specialized Character Generation

CardProjector-v4, developed by WoonaAI, is a 27 billion parameter model based on Google's Gemma 3 architecture, specifically fine-tuned for generating detailed character cards for SillyTavern and general character creation. This version represents a significant advancement in character realism and depth.

Key Capabilities & Features

  • Personality-Centric Generation: Absolute focus on developing deep and realistic character personalities, extensively trained on eight large datasets and MBTI/Enneagram profiles from psychology. It can accurately select personality types based on these systems.
  • Reduced Positive Bias: Trained on a diverse dataset, including aggressive and hypersexualized characters, to significantly reduce inherent positive bias, allowing for the creation of more balanced and realistic negative characters when requested.
  • Gemma 3 Foundation: Transitioned to Gemma 3 for enhanced creative writing skills, higher general knowledge, and improved logical reasoning compared to previous versions.
  • Vision Ability: Supports vision capabilities, allowing for integration with mmproj models (vanilla mmproj for 27B is compatible).
  • Prompt Sensitivity: The model is extremely sensitive to user prompts, executing instructions precisely. Users are advised to be concise yet thorough in their requests.
  • Direct YAML Generation: Recommended for generating characters directly in YAML format for structured and detailed outputs.

Usage Recommendations

  • Chat Template: Gemma
  • Optimal Parameters: Temperature: 0.8, Top-P: 0.92, Rp.Pen: 1.07, Top-K: 100, Rep Pen Range: 360, Rep Pen Slope: 0.7.
  • MBTI/Enneagram Integration: Strongly recommended to include MBTI profiles and Enneagram types for enhanced character depth, either during initial generation or as an add-on.