ComfyAssets-ComfyUI-KikoTools/examples/documentation/text_encode_sampler_params.md

Text Encode Sampler Params

Overview

The Text Encode Sampler Params node combines text encoding with sampler parameter management, providing a unified interface for prompt processing and sampling configuration. Adapted from comfyui-essentials-nodes (now in maintenance mode), this tool streamlines the text-to-image pipeline setup.

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

• Unified Interface: Combine text encoding and sampler params in one node
• Dynamic Prompt Processing: Support for wildcards and syntax
• Parameter Templates: Pre-configured settings for common scenarios
• Batch Text Processing: Handle multiple prompts efficiently
• Model-Aware Encoding: Optimize for different text encoders

Node Properties

• Category: ComfyAssets/🧰 xyz-helpers
• Node Name: TextEncodeSamplerParams
• Function: encode_and_params

Inputs

Required

Parameter	Type	Default	Description
text	STRING	-	Prompt text to encode
clip	CLIP	-	CLIP model for encoding
sampler_name	DROPDOWN	dpmpp_2m	Sampling algorithm
scheduler	DROPDOWN	karras	Noise scheduler
steps	INT	20	Sampling steps
cfg	FLOAT	7.0	CFG scale

Optional

Parameter	Type	Default	Description
negative_text	STRING	""	Negative prompt
seed	INT	-1	Random seed (-1 for random)
denoise	FLOAT	1.0	Denoising strength
template	DROPDOWN	none	Parameter template

Outputs

Name	Type	Description
positive	CONDITIONING	Encoded positive prompt
negative	CONDITIONING	Encoded negative prompt
sampler_params	DICT	Complete sampler parameters

Templates

Portrait Photography

template: "portrait"
→ steps: 25
→ cfg: 7.5
→ sampler: dpmpp_2m_sde
→ scheduler: karras

Landscape Art

template: "landscape"
→ steps: 30
→ cfg: 8.0
→ sampler: dpmpp_3m_sde
→ scheduler: exponential

Quick Preview

template: "preview"
→ steps: 12
→ cfg: 6.0
→ sampler: euler
→ scheduler: normal

High Detail

template: "detailed"
→ steps: 40
→ cfg: 7.0
→ sampler: dpm_adaptive
→ scheduler: karras

Usage Examples

Basic Text-to-Image

TextEncodeSamplerParams → KSampler → VAE Decode
    text: "beautiful landscape"
    negative_text: "ugly, blurry"
    steps: 20

Template-Based Generation

TextEncodeSamplerParams → KSampler
    text: "portrait of a person"
    template: "portrait"
    → Optimized portrait settings

Batch Processing

Multiple Prompts → TextEncodeSamplerParams → Batch Generate
    → Encode all prompts with same settings

Prompt Syntax Support

Wildcards

{red|blue|green} car
→ Randomly selects color

Emphasis

(important:1.2) detail
→ Increases weight to 1.2

Alternation

[cat|dog] in garden
→ Alternates between options

Best Practices

Text Encoding

1. Keep prompts concise and descriptive
2. Use emphasis for important elements
3. Structure prompts logically
4. Test negative prompts impact

Parameter Selection

# Quality over speed
steps: 30-40
cfg: 7-8
sampler: dpmpp_3m_sde

# Speed over quality
steps: 10-15
cfg: 5-6
sampler: euler

Negative Prompts

# Common negatives
"ugly, tiling, poorly drawn, out of frame"

# Style-specific
"cartoon, anime" (for realism)
"realistic, photo" (for artwork)

Integration with Other Nodes

Complete Pipeline

TextEncodeSamplerParams → KSampler → VAE Decode
         ↓                    ↑
    All parameters      From Model Loader

With LoRA

LoRAFolderBatch → TextEncodeSamplerParams → Generate
    → Apply LoRA to encoded text

Multi-Pass Processing

TextEncodeSamplerParams → First Pass (low res)
                       ↘ Second Pass (high res)

Advanced Features

Dynamic Templates

# Load template based on prompt content
if "portrait" in text:
    use_template("portrait")
elif "landscape" in text:
    use_template("landscape")

Prompt Weighting

# Automatic weight calculation
analyze_prompt_importance()
apply_semantic_weights()

CLIP Skip Support

• Adjust CLIP layers used
• Model-specific optimization
• Quality vs style balance

Tips and Tricks

Prompt Optimization

1. Front-load important elements
2. Use commas for separation
3. Avoid contradictions
4. Test with different CFG values

Performance Tuning

# Memory efficient
encode_in_batches = True
clear_cache_between = True

# Speed priority
use_half_precision = True
minimize_conditioning = True

Quality Enhancement

• Higher CFG for prompt adherence
• Lower CFG for creativity
• Balance with step count

Common Workflows

Style Transfer

Reference Image → Extract Style
                ↓
TextEncodeSamplerParams → Apply Style
    text: "in the style of [extracted]"

Prompt Evolution

Base Prompt → Variations → TextEncodeSamplerParams
    → Test different phrasings

A/B Testing

Same prompt → Different parameters → Compare
    template A vs template B

Troubleshooting

Poor Text Adherence

• Increase CFG scale
• Simplify prompt
• Check CLIP model compatibility

Over-saturation

• Reduce CFG scale
• Adjust negative prompt
• Check sampler settings

Encoding Errors

• Verify CLIP model loaded
• Check text formatting
• Remove special characters

Parameter Guidelines

CFG Scale Effects

Low (3-5): Creative, loose interpretation
Medium (6-8): Balanced adherence
High (9-12): Strict prompt following
Very High (13+): Potential artifacts

Step Count Impact

Low (10-15): Fast, rough
Medium (20-30): Good balance
High (40-50): Maximum quality
Very High (50+): Diminishing returns

Model-Specific Settings

SDXL

• CFG: 6-8
• CLIP Skip: 1-2
• Emphasis: Moderate

SD 1.5

• CFG: 7-9
• CLIP Skip: 1-2
• Emphasis: Standard

FLUX

• CFG: 3-5
• CLIP Skip: 0
• Emphasis: Subtle

Version History

• 1.0.0: Initial adaptation from comfyui-essentials-nodes
• 1.0.1: Added template system
• 1.0.2: Enhanced prompt syntax support
• 1.0.3: Improved batch processing

Credits

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