• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
1

Men wearing only wearing suits

Community Beginner ,
Apr 03, 2024 Apr 03, 2024

Copy link to clipboard

Copied

I noticed when I asked the AI to generate an image with a man it only wanted to put the man in a suit! I tried multiple prompts with different clothing items, reference photos of men wearing different articles of clothing but every time it puts them in a suit and tie. Quite funny to be honest, but an issue for sure.

Bug Unresolved
TOPICS
Imaging

Views

56

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
1 Comment
Community Expert ,
Apr 03, 2024 Apr 03, 2024

Copy link to clipboard

Copied

LATEST

"Not wearing suit" is part of your problem. When you want to include "negative" prompts in the main prompt, consider this: the model does not seem to understand the coupling of two words where one word negates the other. Because of that, you end up with the thing you are trying to eliminate actually getting added.  For example,

  • "not wearing suit" ends up with suits
  • "no mountain" ends up generating mountains
  • "without facial hair" ends up showing beards
  • "no men" ends up adding men
  • "no stairs" ends up adding a staircase

 

It is best to try to describe the negative prompt concept in a single word, if possible (even if the word you use is not a real English word):

  • "beardless" rather than "without facial hair"
  • "mountainless" rather than "no mountains"
  • "flat-land" rather than "no mountains"
  • "women" rather than "no men"
  • "stairless" rather than "no stairs"
  • "jacketless shirt-sleeves" rather than "without a suit" (note "suitless" is blocked for some reason)

 

Hyphenating words seems to connect words more than a space, but not as much as if it was a single word (if that makes sense). If two words is a "zero" connexion and one word is a "ten" connexion, hyphenating seems to be around "three" in my experience (if that makes sense). But hypenating does not seem to solve the negative word coupling issue ("no-mountains" is not better than "no mountains").

 

I have no inside information on the model. This is what I have pieced together from using it over the past year. I hope this helps some until the negative prompt box returns.

 

Here is an alternative that follows my guidance above:

droopydog500_0-1712166590663.png

High fashion editorial portrait of jacketless colourful shirt-sleeve funky man 

Votes

Translate

Translate

Report

Report