Exit
  • Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
  • 한국 커뮤니티
0

Exclude terms during generation

Explorer ,
May 30, 2025 May 30, 2025

How can I tell the firefly to NOT produce things? Example: I got 4 pictures. 2 of them with black people. I told firefly to not produce black people. Next round I got 4 out of 4 with black people.

Everybody was wearing a hat or a basecap. I told it not to produce and hats or caps. Guess what happened.

What do I have to promt, to tell the tool NOT to generate something? And why is everyone on my picture looking like an American sitting on a tourist boat? And how can I avoid it?

See wht the fly did and in comparison a single picture in the forst try (from the last prompt) that ChatGTP made!!!Bildschirmfoto-2025-05-30-um-15.12.43.jpg
Bildschirmfoto 2025-05-30 um 15.40.13.png

Idea No status
676
Translate
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

correct answers 1 Correct answer

Community Expert , May 30, 2025 May 30, 2025

Hello @Haggi Hug,

Thank you for your message. 

 

The model does not understand the concept of a word being negated by the word in front of it. So using "not [X]", "no [X]", "exclude [X]", "with no [X]", and "without [X]" does not work. As you observed, it ignores the negation word and adds to the image the thing you are trying to avoid.  You basically need to find "positive" words and phrases telling the model what you want to see rather than telling it what you do not want to see.  Such as "flat

...
Translate
7 Comments
Community Expert ,
May 30, 2025 May 30, 2025

Hello @Haggi Hug,

Thank you for your message. 

 

The model does not understand the concept of a word being negated by the word in front of it. So using "not [X]", "no [X]", "exclude [X]", "with no [X]", and "without [X]" does not work. As you observed, it ignores the negation word and adds to the image the thing you are trying to avoid.  You basically need to find "positive" words and phrases telling the model what you want to see rather than telling it what you do not want to see.  Such as "flat land" rather than "no mountains" or "desolate" rather than "no people".  For people, describe what you want them to be, such as "caucasians" for your example.

 

If I am not sure how to describe something I want to exclude in a positive way, I will ask Google Gemini or ChatGPT for ideas.  Just be aware that when they give prompts, they are usually too wordy and conversational, so you have to edit them, but they are good for ideas.


Thanks,
    droopy

Translate
Report
Explorer ,
May 30, 2025 May 30, 2025

Thank you for your respond. And what can I do to convince the toll to not use people that look like from an USA western or road movie? Or do we just have to admit that firefly is miles behind all other generating tools and poropably will stay there?

Just to leearn: What would you prompt to get a picture like I got from ChatGTP (with the same prompt as I said)? I feel a bit helpless with the poor quality of firefly. To me it feels like you use a outdated tool with poor abillities, compared to other tools (in their free plan, payed they even can do so much more). 😳

Translate
Report
Community Expert ,
Jun 02, 2025 Jun 02, 2025

Hi @Haggi Hug,

 

I could not get rid of the hats, either. Some concepts in Firefly seem to be very ingrained, like beards on men. This might be closer to what you want (except the hats):

droopydog500_0-1748862119702.png

 

 What I would do next, if I could not get rid of something like a hat, is rather than going through dozens of iterations trying to generate an image without them, I would take the images to generative fill and use that to remove the caps.

droopydog500_1-1748862241188.png

to:

droopydog500_2-1748862300937.png

 

Note, even in removing, it took several generations to not get a cap. lol. But, finally got it.  My point is, generative fill can be your friend in fixing what you cannot get the model to do for you in text to image.

Firefly 20250602070725.png

Oh, I just realised these guys do not have tattoos because of heavy clothing. I did have them with tattoos earlier. I could tweek my prompt to change the clothing or use generative fill to change the clothing and add tattoos.

 

My best,

    droopy

Translate
Report
Explorer ,
Jun 02, 2025 Jun 02, 2025

Thank you very much for the work. I will think on it. But as you already noticed now the sailors have no tatoos and because of whatever reason it’s cold outside, so they wear thick jackets. I think the fly now took „caucasian“ and placed the boat somwhere near russia. To be honest, this is not something you can work profesionally with. Not at all. Other generators are doing so much better and it doesn’t seem that Adobe is catching up.
(By the way: the sepia tone Chat GPT did by its own, I would appreciate to have something in natural colours.)

Example from Leonardo:
Bildschirmfoto 2025-06-02 um 15.30.24.png

Translate
Report
Community Expert ,
Jun 02, 2025 Jun 02, 2025

Oh, ok. I did sepia because it was in the image you shared from ChatGPT.

 

I changed the beginning of the promt to "two heavily tattoo-covered shirtless caucasian sailors in hot Carribean" and the heavy coats disappeared and the tattoos were back.

 

I understand your concerns about the quality and encouage you to thumbs down any images you feel are poor quality. The models are constantly changing and improving.

 

My best,

    droopy

Translate
Report
Explorer ,
Jun 02, 2025 Jun 02, 2025

Yes, I know. But I have no time to play around with Adobe’s software. I will come again later and test it after some months. In the moment you can forget it to produce professional jobs. I could write a book about all the useless features and not functioning interfaces they have. They always are excited to produce new gimmicks but they never seem to work with it. And the more they invent, the less consistent everything becomes. But they are so happy that they raise the prices. Yeah. Annoying.

 

When it comes to picture generation they really are on the last places.

Translate
Report
Adobe Employee ,
Jun 02, 2025 Jun 02, 2025
LATEST

Thank you for taking the time to share your thoughts. Your insights into interface usability and feature consistency are valuable to us, and we are committed to improving stability while innovating thoughtfully. If there are specific issues affecting your experience, we’d love to hear more so we can pass them along to our team.


^KR

Translate
Report