Model Selection Guide

This is the authoritative reference for AI models, usage limits, and capabilities in NPC Forge. All usage information in other documentation files references this guide.

Model Overview

NPC Forge uses two types of AI models with three quality tiers each:

Text Models: Generate character attributes, quests, dialogue, items, chat responses, and traits
Image Models: Create and edit character portraits

Each tier offers different quality levels and monthly usage limits, allowing you to balance output quality with available generations.

Text Generation Models

🟢 Standard Tier: gpt-4o-mini

Monthly Limit: 50 generations
Reset Date: 1st of each month
Quality: Good baseline quality suitable for most characters
Best For:
Regular character creation and experimentation
Background NPCs and supporting characters
Casual chat conversations
Frequent trait generation and editing
Testing different character concepts

🟡 Enhanced Tier: gpt-4.1-mini

Monthly Limit: 30 generations
Reset Date: 1st of each month
Quality: Improved detail, narrative consistency, and character depth
Best For:
Important story characters requiring depth
Meaningful chat conversations about plot or relationships
Higher quality quest and dialogue generation
Characters central to your campaign or story
Targeted trait development for key NPCs

🔴 Premium Tier: gpt-4o

Monthly Limit: 10 generations
Reset Date: 1st of each month
Quality: Highest quality output with rich detail and sophisticated personality development
Best For:
Critical main characters and primary antagonists
Key story moments and dramatic chat conversations
Maximum detail character creation
Premium trait development for complex personalities
Characters requiring nuanced psychological depth

Image Generation Models

🟢 Standard Tier: DALL-E 2

Monthly Limit: 10 generations
Reset Date: 1st of each month
Quality: Good basic portrait generation with clear character representation
Portrait Editing: Not supported - no editing capabilities
Best For:
Quick character visualization
Background NPCs requiring portraits
Testing portrait concepts

🟡 Enhanced Tier: DALL-E 3

Monthly Limit: 5 generations
Reset Date: 1st of each month
Quality: Improved detail, artistic quality, and visual consistency
Portrait Editing: Not supported - no editing capabilities
Best For:
Important character portraits requiring quality
Characters with complex visual descriptions
Final portraits for key NPCs

🔴 Premium Tier: gpt-image-1

Monthly Limit: 3 generations
Reset Date: 1st of each month
Quality: Maximum artistic control, detail, and visual fidelity
Portrait Editing: Full editing support with best results
Best For:
Showcase-quality character portraits
Characters requiring portrait editing capabilities
Complex visual concepts needing highest quality
Final portraits for main characters

Portrait Editing Capabilities

Model Support for Portrait Editing

gpt-image-1 (Premium) - ONLY MODEL SUPPORTING EDITING

Full editing support with reliable, high-quality results
Handles complex edits: color changes, accessories, clothing modifications, expression changes
Edit prompt limit: Up to 32,000 characters for detailed instructions
Best quality editing with consistent results
Supports all edit types: additions, removals, color changes, style modifications

dall-e-2 (Standard) - NO EDITING SUPPORT

Cannot be used for portrait editing operations
Generate-only model for initial portrait creation
No modification capabilities available
Must regenerate entirely for changes

dall-e-3 (Enhanced) - NO EDITING SUPPORT

Cannot be used for portrait editing operations
Generate-only model for initial portrait creation
No modification capabilities available
Must regenerate entirely for changes

Portrait Editing Use Cases

Successful Edit Types:

Color Modifications: Hair color, eye color, clothing colors, skin tone adjustments
Accessory Management: Adding/removing hats, glasses, jewelry, weapons, armor
Expression Changes: Smile adjustments, serious expressions, emotional modifications
Clothing Updates: Outfit changes, adding/removing layers, style modifications
Detail Enhancement: Scars, tattoos, facial hair, age modifications

Edit Prompt Examples:

"change hair color to blonde"
"add round glasses"
"remove the hat"
"make them smile"
"change shirt to red"
"add a beard"
"give them a serious expression"
"add leather armor"

Chat Integration

How Chat Uses Text Models

Chat conversations utilize the selected text model for generating character responses:

Each chat response counts as one generation against your monthly text model limit
Model selection can be changed during conversations
Response quality directly correlates with selected model tier
Character personality consistency improves with higher tiers

Chat-Specific Model Performance

Standard Model (gpt-4o-mini) for Chat:

Response Quality: Clear, coherent character responses with basic personality consistency
Character Voice: Adequate for general character interactions and world-building
Context Understanding: Good grasp of immediate conversation context
Best For: Daily character interactions, casual conversations, background character dialogue

Enhanced Model (gpt-4.1-mini) for Chat:

Response Quality: More nuanced character voice with improved personality depth
Character Voice: Better consistency and understanding of character relationships
Context Understanding: Superior comprehension of character motivations and background
Best For: Important character development conversations, plot discussions, relationship building

Premium Model (gpt-4o) for Chat:

Response Quality: Sophisticated character development with excellent personality consistency
Character Voice: Most natural and engaging character interactions
Context Understanding: Deep understanding of character relationships, history, and motivations
Best For: Critical story moments, emotional scenes, main character interactions, complex plot discussions

Chat Usage Patterns

Conversation Planning:

Daily Interactions (Standard): Greetings, small talk, general world-building
Character Development (Enhanced): Exploring backstory, developing relationships, important discussions
Key Story Moments (Premium): Major revelations, emotional scenes, critical decision points

Trait Management Integration

Text Models for Trait Generation

Trait Generation Capabilities:

"Add Generated Trait" counts as one text generation against monthly limits
Individual trait regeneration counts as one generation per trait
Multiple traits generated simultaneously count as one generation
Trait quality improves with higher model tiers

Model Performance for Traits:

Standard Model:

Quality: Good basic trait generation appropriate to character context
Consistency: Adequate integration with existing character elements
Best For: Frequent trait additions, experimental character development

Enhanced Model:

Quality: Better trait quality with improved character consistency
Consistency: Superior integration with character personality and background
Best For: Important character trait development, refined character elements

Premium Model:

Quality: Highest quality trait generation with nuanced character understanding
Consistency: Excellent contextual appropriateness and character integration
Best For: Critical character traits, complex personality development, main character enhancement

Usage Tracking and Management

Monthly Limit System

Limit Tracking:

Individual tracking for each model tier (text and image separately)
Monthly reset on the 1st of each month at 00:00 UTC
Real-time usage display in model selection interfaces
Warning notifications when approaching limits

Usage Calculation:

Character Generation: Counts against selected text and image models
Chat Responses: Count against selected text model only
Portrait Editing: Counts against selected image model only
Trait Operations: Count against selected text model only
Regeneration: Each operation counts as one generation

Development Mode:

Usage limits bypassed in development environment for testing
Production deployment enforces all limits normally
Local development can test all features without limit restrictions

Strategic Usage Planning

Efficient Usage Strategies:

Model Combination Approaches:

Standard Text + Enhanced Image: Good balance for most characters
Enhanced Text + Standard Image: Better character depth with basic visualization
Premium Text + Premium Image: Maximum character development with editing capabilities
Mixed Regeneration: Start with Standard, upgrade specific elements with higher tiers

Monthly Planning:

Early Month: Experiment with Standard models, create base characters
Mid-Month: Use Enhanced models for important character development
End-Month: Reserve Premium models for critical characters and key moments

Feature-Specific Strategies:

Character Creation: Use Standard for testing, Enhanced/Premium for final versions
Chat Conversations: Start with Standard, upgrade for important story moments
Portrait Editing: Plan edits carefully due to limited Premium image generations
Trait Management: Use Standard for frequent additions, Enhanced/Premium for key character traits

Usage Limit Comparison Table

Feature	Standard Monthly	Enhanced Monthly	Premium Monthly
Text Model Limit	50 generations	30 generations	10 generations
Image Model Limit	10 generations	5 generations	3 generations
Character Creation	✅ Unlimited variety	✅ Better quality	✅ Highest quality
Chat Conversations	✅ Casual interactions	✅ Character development	✅ Key story moments
Portrait Editing	❌ Not supported	❌ Not supported	✅ Full support
Trait Generation	✅ Frequent additions	✅ Quality improvements	✅ Premium development
Best Use Case	Regular characters	Important characters	Critical characters

Model Selection Recommendations

For New Users

1Start with Standard models to learn the system and experiment
2Create several test characters to understand quality differences
3Try different combinations of text and image model tiers
4Gradually use Enhanced/Premium for characters you want to keep

For Regular Users

1Use Standard models for background NPCs and experimentation
2Reserve Enhanced models for important story characters
3Save Premium models for main characters and critical moments
4Plan usage around monthly cycles for optimal distribution

For Campaign Managers

1Standard Tier: Shopkeepers, guards, random encounters
2Enhanced Tier: Recurring NPCs, faction leaders, important allies
3Premium Tier: Main antagonists, key story characters, party mentors

For Writers and Content Creators

1Standard Tier: Background characters, world-building NPCs
2Enhanced Tier: Secondary characters, supporting cast
3Premium Tier: Protagonists, primary antagonists, central figures

Quality Comparison Examples

Text Generation Quality Differences

Character Description Generation:

Standard: "Marcus is a thoughtful monk who values knowledge and questions authority."
Enhanced: "Marcus embodies scholarly curiosity beneath a humble exterior, his gentle demeanor concealing profound theological questions that challenge orthodox doctrine."
Premium: "Brother Marcus moves through the monastery with deliberate quietude, his ink-stained fingers betraying countless hours spent not merely copying sacred texts, but wrestling with their deeper implications—a scholar whose faith is deepened rather than diminished by intellectual honesty."

Image Generation Quality Differences

Portrait Generation Results:

Standard (DALL-E 2): Clear, recognizable portraits with basic detail and good color accuracy
Enhanced (DALL-E 3): Improved artistic quality, better lighting, more sophisticated composition and detail
Premium (gpt-image-1): Maximum visual fidelity, superior artistic control, exceptional detail and style consistency

Technical Considerations

Model Performance Characteristics

Response Time Expectations:

Text Generation: 10-30 seconds for character content
Image Generation: 30-120 seconds depending on model and complexity
Chat Responses: 5-15 seconds for conversational responses
Portrait Editing: 30-90 seconds for image modifications

Model Availability:

All models available 24/7 through OpenAI infrastructure
Temporary slowdowns possible during high-demand periods
Automatic retry logic for transient failures
Error handling with user-friendly feedback

Integration Details

API Integration:

Server-side processing for all model interactions
Secure API key handling without client exposure
Usage validation before expensive operations
Error categorization for specific user guidance