Gemma-3-Starshine-12B Open-Source Creative Writing Model - Freely Assist in Story Creation and Scene Description

Gemma 3 Starshine 12B

Developed by ToastyPigeon

A creative writing model fine-tuned and merged from Gemma 3 12B IT and Gemma 3 12B PT, specializing in story creation and scene description

Large Language Model

Transformers

#Creative Writing #Novel Style #Story-Oriented

Downloads 329

Release Time : 3/27/2025

Model Overview

This is a story-oriented merged version with a novel-like writing style, excelling at imitating user characters for creative writing. The model combines instruction-following capabilities with free-style writing features, making it suitable for storytelling and scene description.

Model Features

Story-Oriented Writing

Optimized for creative writing and storytelling with a novel-like style

Character Imitation Ability

Excels at imitating user characters for text generation

Hybrid Model Advantages

Combines the strengths of instruction models and base models, featuring both instruction-following capability and free-style writing

Visual Tower Functionality

Supports vision-related text generation capabilities

Model Capabilities

Creative text generation

Storytelling

Scene description

Character imitation

Instruction following

Use Cases

Creative Writing

Novel Creation

Generates coherent novel chapters and plots

Fluid narrative structure and vivid scene descriptions

Role-Playing

Imitates specific character styles for dialogue and narration

Consistent personalized text output in character style

Edutainment

Interactive Story Creation

Collaborates with users to create stories

Dynamically responds to user input with creative text

🌠 G3 Starshine 12B 🌠

This is a creative writing model that combines fine - tuned versions of Gemma 3 12B IT and Gemma 3 12B PT. It focuses on storytelling and scenarios.

This was Merge A / A1 in the testing set.

A creative writing model is based on a merge of fine - tunes on Gemma 3 12B IT and Gemma 3 12B PT. This is the Story Focused merge. This version performs better for storytelling and scenarios, as the prose is more novel - like and it has a tendency to impersonate the user character.

Check out the Alternate RP Focused version too.

This is a merge of two G3 models, one trained on instruct and one trained on base:

allura-org/Gemma-3-Glitter-12B - Itself a merge of a storywriting and RP train (both also by ToastyPigeon), on instruct
ToastyPigeon/Gemma-3-Confetti-12B - Experimental application of the Glitter data using base instead of instruct, additionally includes some adventure data in the form of SpringDragon.

The result is a lovely blend of Glitter's ability to follow instructions and Confetti's free - spirit prose, effectively 'loosening up' much of the hesitancy that was left in Glitter.

Update: Vision tower is back! Have fun.

Thank you to jebcarter for the idea to make this. I love how it turned out!

🚀 Quick Start

Instruct Format

This model uses Gemma2/3 instruct and has been trained to recognize an optional system role.

⚠️ Important Note

While it won't immediately reject the system role, the results may be better without it.

<start_of_turn>system
{optional system turn with prompt}<end_of_turn>
<start_of_turn>user
{User messages; can also put sysprompt here to use the built-in g3 training}<end_of_turn>
<start_of_turn>model
{model response}<end_of_turn>

Merge Configuration

Surprisingly, after trying several methods, this one worked best.

models:
  - model: ToastyPigeon/Gemma-3-Confetti-12B
    parameters:
      weight: 0.5
  - model: allura-org/Gemma-3-Glitter-12B
    parameters:
      weight: 0.5
merge_method: linear
tokenizer_source: allura-org/Gemma-3-Glitter-12B

📚 Documentation

Model Information

Property	Details
Base Model	allura-org/Gemma-3-Glitter-12B, ToastyPigeon/Gemma-3-Confetti-12B, google/gemma-3-12b-it, google/gemma-3-12b-pt
Library Name	transformers
Tags	mergekit, merge

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご