Negative Prompt Examples: 120 Copy/Paste Lines That Actually Work

Negative prompts are not “extra words”. They’re guardrails.

I used to paste random negatives like “ugly, bad hands” and hope for the best. It kinda worked… until it didn’t. The fix was treating negatives like a QA checklist: remove recurring failure modes (text, logos, anatomy glitches, CGI skin, compression artifacts).

If you want the full framework (formula + Prompt Kit): start here → prompt engineering for image models.

Baseline negative prompt (use everywhere)

This is the baseline I start with for most photoreal prompts. Then I add categories only when a specific issue keeps showing up.

text, watermark, logo, signature,
blurry, low-res, jpeg artifacts,
deformed hands, extra fingers, missing fingers, bad anatomy, warped face,
plastic skin, oversmoothed, overprocessed, cartoon, CGI

How to use this list (without overkilling your prompt)

1. Start with the baseline.
2. If you still get an issue (e.g., text), add the relevant category lines.
3. Don’t paste all 120 negatives unless you need them — it can reduce variety.

120 negative prompt examples (copy/paste)

Text / logos / watermarks

1) text
2) watermark
3) logo
4) signature
5) brand mark
6) caption
7) subtitles
8) UI text
9) letters
10) numbers
11) typography
12) gibberish text
13) random words
14) label text
15) sticker
16) stamp
17) QR code
18) barcode
19) website URL
20) @handle
21) social media overlay
22) copyright mark
23) trademark symbol
24) product packaging text
25) poster text
26) menu text

Anatomy (hands / faces / proportions)

27) deformed hands
28) extra fingers
29) missing fingers
30) fused fingers
31) bad anatomy
32) warped face
33) asymmetrical face
34) duplicate face
35) cross-eyed
36) misaligned eyes
37) bad teeth
38) extra limbs
39) missing limbs
40) broken joints
41) elongated neck
42) twisted arms
43) mutated fingers
44) melted hands
45) melted face
46) two heads
47) duplicate body
48) incorrect proportions
49) bad feet
50) extra toes
51) missing toes
52) deformed ears
53) crooked jaw
54) blown-out pupils
55) uncanny eyes
56) doll-like eyes
57) overly glossy eyes
58) misplaced eyebrows
59) blurred facial features
60) smudged face
61) bad skin texture
62) waxy skin
63) plastic skin
64) oversmoothed skin
65) airbrushed skin
66) poreless skin
67) low-detail skin
68) muddy skin
69) weird skin
70) skin artifacts
71) banding on skin
72) texture stretching
73) distorted mouth
74) distorted nose
75) distorted ears
76) bad hairline

Style drift (CGI / cartoon / illustration)

77) cartoon
78) anime
79) 3d
80) cgi
81) render
82) illustration
83) digital painting
84) oil painting
85) watercolor
86) comic
87) toon
88) pixar style
89) plastic look
90) toy-like
91) doll
92) overprocessed
93) over-sharpened
94) HDR
95) oversaturated
96) posterized
97) low contrast
98) flat lighting
99) fake bokeh
100) excessive bokeh
101) lens distortion
102) fisheye
103) wide distortion

Quality / camera artifacts

104) blurry
105) out of focus
106) low-res
107) low resolution
108) jpeg artifacts
109) noise
110) grainy (bad)
111) pixelated
112) compression artifacts
113) motion blur
114) ghosting
115) double exposure
116) chromatic aberration
117) color fringing
118) bad lighting
119) underexposed
120) overexposed
121) harsh flash
122) blown highlights
123) crushed blacks
124) banding
125) moiré
126) aliasing

FAQ

Should I always use negative prompts?

Yes — at least the baseline. Expand only when needed.

Why do I still get text?

Use negatives (text/watermark/logo) and also add constraints in your main prompt: “no text, no watermark”.

Do negatives make images worse sometimes?

If you paste a massive list, you can get bland results. Keep it targeted.

Your mission

Run 10 generations with the baseline. If one issue repeats (hands/text/CGI), add only that category. Iterate like a producer, not like a gambler.

If you want the full creative workflow: Sistema Criativo: Diretor de Arte IA (Portuguese audio + Hotmart subtitles).