prompt-engineering

1# ComfyUI Prompt Engineering
2 
3## CLIP Text Encoding Fundamentals
4 
5ComfyUI uses CLIP (Contrastive Language-Image Pre-training) text encoders to convert text prompts into conditioning tensors. The `CLIPTextEncode` node takes a text string and a CLIP model, producing a `CONDITIONING` output for the KSampler.
6 
7### Token Limit
8 
9CLIP processes text in **77-token chunks**. Each word is typically 1-3 tokens. Prompts exceeding 77 tokens are silently truncated unless you use the BREAK token or a multi-clip encoding node.
10 
11## Weight Syntax
12 
13### Emphasis (Attention Weights)
14 
15Adjust how strongly the model attends to specific words or phrases:
16 
17| Syntax | Effect | Equivalent Weight |
18|--------|--------|-------------------|
19| `(word:1.3)` | Increase emphasis by 30% | Explicit weight 1.3 |
20| `(word:0.7)` | Decrease emphasis by 30% | Explicit weight 0.7 |
21| `(word)` | Slight increase | `(word:1.1)` |
22| `((word))` | Moderate increase | `(word:1.21)` — 1.1^2 |
23| `(((word)))` | Strong increase | `(word:1.331)` — 1.1^3 |
24| `[word]` | Slight decrease | `(word:0.9091)` — 1/1.1 |
25| `[[word]]` | Moderate decrease | `(word:0.8264)` — 1/1.1^2 |
26 
27### Weight Rules
28 
29- **Valid range**: 0.0 to 2.0 (going beyond 1.5 often causes artifacts)
30- **Default weight**: 1.0 for unmodified tokens
31- **Nesting stacks multiplicatively**: `((word))` = 1.1 * 1.1 = `(word:1.21)`
32- **Phrases**: `(red sports car:1.3)` applies weight to the entire phrase
33…