nodes/ComfyAssets-ComfyUI-KikoTools

Fork 0

Files

T

Vito Sansevero f2093d6567 style(docs): Ensure newline at EOF in markdown files

2025-12-17 07:01:22 -08:00

6.6 KiB

Raw Permalink Blame History

Sampler Select Helper

Overview

The Sampler Select Helper node provides intelligent sampler selection with model-specific recommendations and compatibility checking. Adapted from comfyui-essentials-nodes (now in maintenance mode), this tool ensures optimal sampler-scheduler combinations for different model architectures.

Attribution

This node is based on work from comfyui-essentials-nodes by cubiq. The original project is in maintenance-only mode, and we've adopted and enhanced these tools to ensure continued support and compatibility with modern ComfyUI workflows.

Features

Model-Aware Selection: Automatic recommendations based on model type
Compatibility Validation: Ensures sampler-scheduler pairs work well together
Performance Profiles: Pre-configured settings for quality vs speed
Dynamic Updates: Adapts to newly available samplers
Batch Support: Test multiple samplers in sequence

Node Properties

Category: ComfyAssets/🧰 xyz-helpers
Node Name: SamplerSelectHelper
Function: select_sampler

Inputs

Required

Parameter	Type	Default	Description
`model_type`	DROPDOWN	auto	[auto, sdxl, sd15, flux, custom]
`quality_preset`	DROPDOWN	balanced	[fast, balanced, quality, extreme]
`sampler_override`	DROPDOWN	auto	Specific sampler selection

Optional

Parameter	Type	Default	Description
`scheduler_override`	DROPDOWN	auto	Specific scheduler selection
`model_name`	STRING	-	Model name for auto-detection
`custom_rules`	STRING	-	JSON rules for custom selection

Outputs

Name	Type	Description
`sampler_name`	STRING	Selected sampler
`scheduler`	STRING	Selected scheduler
`recommended_steps`	INT	Suggested step count
`recommended_cfg`	FLOAT	Suggested CFG scale

Model-Specific Recommendations

SDXL Models

quality_preset: "balanced"
→ sampler: "dpmpp_2m"
→ scheduler: "karras"
→ steps: 25
→ cfg: 7.0

SD 1.5 Models

quality_preset: "quality"
→ sampler: "dpmpp_2m_sde"
→ scheduler: "exponential"
→ steps: 30
→ cfg: 7.5

FLUX Models

quality_preset: "fast"
→ sampler: "euler"
→ scheduler: "simple"
→ steps: 15
→ cfg: 3.5

Quality Presets Explained

Fast (Preview)

Goal: Quick iterations
Steps: 10-15
Samplers: euler, dpm_fast
Use Case: Testing prompts

Balanced (Default)

Goal: Good quality/speed ratio
Steps: 20-25
Samplers: dpmpp_2m, dpmpp_2m_sde
Use Case: Regular generation

Quality

Goal: Best visual quality
Steps: 30-40
Samplers: dpmpp_3m_sde, dpm_adaptive
Use Case: Final renders

Extreme

Goal: Maximum quality
Steps: 50-100
Samplers: dpm_adaptive, dpmpp_3m_sde
Use Case: Hero images

Usage Examples

Auto Model Detection

Load Model → SamplerSelectHelper → KSampler
    model_type: auto
    quality_preset: balanced

Custom Override

SamplerSelectHelper → KSampler
    sampler_override: "dpmpp_3m_sde"
    scheduler_override: "exponential"

Batch Testing

SamplerSelectHelper → Batch Process
    quality_preset: [fast, balanced, quality]
    → Compare outputs

Compatibility Matrix

Recommended Combinations

Sampler	Best Schedulers	Avoid
euler	normal, karras	sgm_uniform
euler_a	normal, karras	simple
dpmpp_2m	karras, exponential	-
dpmpp_2m_sde	karras, exponential	simple
dpmpp_3m_sde	exponential	simple
dpm_adaptive	normal	karras

Best Practices

Model Type Detection

Use auto for automatic detection
Override only when necessary
Provide model_name for better accuracy

Performance Optimization

# Quick preview workflow
quality_preset: "fast"
→ 10 steps, euler sampler

# Final production
quality_preset: "quality"
→ 35 steps, dpmpp_3m_sde

# Experimental/artistic
quality_preset: "extreme"
→ 75 steps, dpm_adaptive

Custom Rules Format

{
  "model_pattern": "anime.*",
  "sampler": "dpmpp_2m_sde",
  "scheduler": "karras",
  "steps": 28,
  "cfg": 7.0
}

Integration with Other Nodes

Complete Pipeline

Model Loader → SamplerSelectHelper → KSampler
             ↘ FluxSamplerParams ↗

A/B Testing

SamplerSelectHelper → KSampler → Image A
    quality: fast
SamplerSelectHelper → KSampler → Image B
    quality: quality
→ Compare Results

Advanced Features

Dynamic Sampler Discovery

Automatically detects new samplers
Updates compatibility matrix
Maintains optimal pairings

Performance Profiling

Tracks generation times
Suggests optimal settings
Adapts to hardware capabilities

Model Fingerprinting

Identifies model architecture
Applies specific optimizations
Learns from usage patterns

Tips and Tricks

Speed vs Quality

Start with "fast" for prompt testing
Move to "balanced" for iteration
Use "quality" for final output
Reserve "extreme" for special cases

Sampler Selection Logic

if model_type == "flux":
    prefer ["euler", "dpmpp_2m"]
elif model_type == "sdxl":
    prefer ["dpmpp_2m_sde", "dpmpp_3m_sde"]
else:
    use ["dpmpp_2m", "euler_a"]

Memory Considerations

Fast presets use less memory
Extreme presets may require more VRAM
Adaptive samplers adjust dynamically

Troubleshooting

Wrong Sampler Selected

Check model_type setting
Verify model detection
Use manual override if needed

Poor Quality Output

Increase quality preset
Check recommended steps
Verify CFG scale

Performance Issues

Start with fast preset
Reduce step count
Try simpler samplers

Common Workflows

Model Comparison

Test same prompt across different models with optimal settings for each.

Quality Ladder

Progress from fast to extreme quality to find optimal balance.

Sampler Shootout

Compare all compatible samplers for specific model/prompt combination.

Version History

1.0.0: Initial adaptation from comfyui-essentials-nodes
1.0.1: Added FLUX model support
1.0.2: Enhanced compatibility matrix
1.0.3: Improved auto-detection

Credits

Original implementation by cubiq in comfyui-essentials-nodes. Adapted and maintained by the ComfyAssets team.

6.6 KiB Raw Permalink Blame History