Template Documentation

Overview

This document provides comprehensive documentation for all templates available in the MCP Server for Replicate. Templates are organized into several categories:

1. Model Parameters
2. Common Configurations
3. Prompt Templates

Model Parameters

SDXL Parameters

The SDXL template provides parameters optimized for Stable Diffusion XL models.

{
    "prompt": "your detailed prompt",
    "negative_prompt": "elements to avoid",
    "width": 1024,  # 512-2048, multiple of 8
    "height": 1024,  # 512-2048, multiple of 8
    "num_inference_steps": 50,  # 1-150
    "guidance_scale": 7.5,  # 1-20
    "prompt_strength": 1.0,  # 0-1
    "refine": "expert_ensemble_refiner",  # or "no_refiner", "base_image_refiner"
    "scheduler": "K_EULER",  # or "DDIM", "DPM_MULTISTEP", "PNDM", "KLMS"
    "num_outputs": 1,  # 1-4
    "high_noise_frac": 0.8,  # 0-1
    "seed": null,  # null or integer
    "apply_watermark": true
}

SD 1.5 Parameters

The SD 1.5 template provides parameters optimized for Stable Diffusion 1.5 models.

{
    "prompt": "your detailed prompt",
    "negative_prompt": "elements to avoid",
    "width": 512,  # 256-1024, multiple of 8
    "height": 512,  # 256-1024, multiple of 8
    "num_inference_steps": 50,  # 1-150
    "guidance_scale": 7.5,  # 1-20
    "scheduler": "K_EULER",  # or "DDIM", "DPM_MULTISTEP", "PNDM", "KLMS"
    "num_outputs": 1,  # 1-4
    "seed": null,  # null or integer
    "apply_watermark": true
}

ControlNet Parameters

The ControlNet template provides parameters for controlled image generation.

{
    "control_image": "image_url_or_base64",
    "control_mode": "balanced",  # or "prompt", "control"
    "control_scale": 0.9,  # 0-2
    "begin_control_step": 0.0,  # 0-1
    "end_control_step": 1.0,  # 0-1
    "detection_resolution": 512,  # 256-1024, multiple of 8
    "image_resolution": 512,  # 256-1024, multiple of 8
    "guess_mode": false,
    "preprocessor": "canny"  # or other preprocessors
}

Common Configurations

Quality Presets

Pre-configured quality settings for different use cases:

• draft: Fast iterations (20 steps)
• balanced: General use (30 steps)
• quality: High quality (50 steps)
• extreme: Maximum quality (150 steps)

Style Presets

Pre-configured style settings:

• photorealistic: Highly detailed photo style
• cinematic: Movie-like dramatic style
• anime: Anime/manga style
• digital_art: Modern digital art style
• oil_painting: Classical painting style

Aspect Ratio Presets

Common aspect ratios with optimal resolutions:

• square: 1:1 (1024x1024)
• portrait: 2:3 (832x1216)
• landscape: 3:2 (1216x832)
• wide: 16:9 (1344x768)
• mobile: 9:16 (768x1344)

Negative Prompt Presets

Quality control negative prompts:

• quality_control: Basic quality control
• strict_quality: Comprehensive quality control
• photo_quality: Photo-specific quality control
• artistic_quality: Art-specific quality control

Prompt Templates

Text-to-Image

Detailed Scene Template

{subject} in {setting}, {lighting} lighting, {mood} atmosphere, {style} style, {details}

Example:

"a young explorer in ancient temple ruins, dramatic golden hour lighting, mysterious atmosphere, cinematic style, vines growing on weathered stone, dust particles in light beams"

Character Portrait Template

{gender} {character_type}, {appearance}, {clothing}, {expression}, {pose}, {style} style, {background}

Landscape Template

{environment} landscape, {time_of_day}, {weather}, {features}, {style} style, {mood} mood

Image-to-Image

Style Transfer Template

Transform into {style} style, {quality} quality, maintain {preserve} from original

Variation Template

Similar to original but with {changes}, {style} style, {quality} quality

ControlNet

Pose-Guided Template

{subject} in {pose_description}, {clothing}, {style} style, {background}

Depth-Guided Template

{subject} with {depth_elements}, {perspective}, {style} style

Best Practices

1. Parameter Selection

   • Start with preset configurations
   • Adjust parameters gradually
   • Use appropriate aspect ratios for your use case

2. Prompt Engineering

   • Use detailed, specific descriptions
   • Include style and quality indicators
   • Use negative prompts for quality control

3. ControlNet Usage

   • Match detection and output resolutions
   • Use appropriate preprocessors for your use case
   • Adjust control scale based on desired influence

4. Quality Optimization

   • Use higher step counts for final outputs
   • Adjust guidance scale for creativity vs. accuracy
   • Use refiners for enhanced quality

Version History

• v1.1.0: Added comprehensive parameter descriptions and validation
• v1.0.0: Initial release with basic parameters