SD Prompt Debugger

SD prompt debugger

When a Stable Diffusion prompt produces the wrong thing, the cause is usually mechanical, not mysterious: the prompt overflows 75 tokens, the weight syntax is malformed, two style terms conflict, or the negative prompt quietly cancels your subject. This debugger checks all of those and tells you what to change.

How it works

The tool runs several deterministic checks:

token estimate      → warn if > 75 (CLIP chunk boundary)
bracket matching    → unbalanced ( ) or [ ] and missing weight colons
weight range check  → flag (term:>1.5) which tends to "fry" images
conflict scan       → opposing style/time/shot terms in one prompt
quality boosters    → note if none present (model-dependent)
neg vs pos overlap  → words appearing in both prompts

Token counting is an estimate (roughly words plus punctuation), which is close enough to catch the cases that matter. Quality-booster advice is tuned to the model version you select.

Tips and common fixes

• Front-load the subject. Keep the main subject and key descriptors inside the first ~60 tokens so they land in the first CLIP chunk.
• Use sane weights. Stay within roughly 0.6–1.4; weights above ~1.5 distort.
• Resolve conflicts. Don’t ask for “minimalist” and “highly detailed” at once unless you deliberately weight one down.
• Check the negative prompt. A word like “dark” in the negative will suppress a deliberately dark scene — remove overlaps.
• SDXL needs fewer boosters. Tags like “masterpiece, best quality” help SD 1.5 but matter far less on SDXL, which is trained on cleaner captions.