CLIPTextEncode Node

Encode text prompts for image generation.

Function

Converts text into token embeddings that the model can understand.

Inputs

Input	Type	Description
clip	clip	CLIP model from CheckpointLoader
text	string	Prompt text

Outputs

Output	Type	Description
CONDITIONING	conditioning	Encoded prompt for KSampler

Parameters

Parameter	Type	Description
text	text	Prompt or negative prompt

Usage

Positive Prompt

CheckpointLoader
    ↓ CLIP
CLIPTextEncode ("a beautiful landscape")
    ↓ CONDITIONING
KSampler (positive)

Negative Prompt

CheckpointLoader
    ↓ CLIP
CLIPTextEncode ("blurry, low quality")
    ↓ CONDITIONING
KSampler (negative)

SDXL Usage

SDXL uses two CLIP encoders:

CheckpointLoader
    ↓ CLIP (1)
CLIPTextEncodeSDXL → KSampler
    ↓ CLIP (2)
CLIPTextEncodeSDXL (same text)

Prompt Syntax

Quality Tags

# Add quality tags
masterpiece, best quality, ultra detailed

# Use parentheses for emphasis
(red hair) = more important
[(blue eyes)] = less important

Weights

# Increase weight
(blue eyes:1.3)  # 30% more

# Decrease weight
(background:0.8)  # 20% less

Tips

1. Be specific: More detail = better results
2. Quality tags: Add "masterpiece, best quality"
3. Negative prompts: Exclude unwanted elements

Common Prompts

Positive

masterpiece, best quality, detailed, 8k

Negative

blurry, low quality, distorted, ugly

Next Steps

• KSampler - Use conditioning
• Checkpoint - Load model