Skip to main content
Participant
March 24, 2023
Released

P: Negative prompting

  • March 24, 2023
  • 70 replies
  • 31045 views

In the text to image part, I didn't find a place to cancel prompts like in stable diffusion. Often we have artifacts with double faces or eyes in the wrong places. It depends on the channels I think but it could be efficient and necessary to add this function. Or maybe add a basic lexicon for prompts if we have to use No in front of cancels or others basic operations.

 

Thank you for your job, im a teacher and i think you are the best to create an democratic useful app for commons users

 

Eric

70 replies

jane-e
Community Expert
Community Expert
September 8, 2024

 

 

@Anitaphotography.nl wrote: There is literally zero info about negative prompts in that link.

 

 

 

Hi Anita

Please see the reply marked Correct from Staff at the top of the page. Negative prompts have been removed and are on pause.

Jane

 

Anitaphotography.nl
Participating Frequently
September 8, 2024

There is literally zero info about negative prompts in that link.

jane-e
Community Expert
Community Expert
September 1, 2024

 

@Yunus Emre37159996nzdb wrote:

This is simply outrageous, how come open source projects have better functionality than this paid alternative?

 

 

Do you need assistance with Adobe Firefly? If so, please provide details and screenshots so we can better assist.

 

Jane

Participant
September 1, 2024

This is simply outrageous, how come open source projects have better functionality than this paid alternative?

The model is completely lobotomized to miss the art style of the reference image the moment I decide to introduce any term that describes a person and mix races along with it.

The moment I get the hardware you WILL lose a customer.

Participant
August 29, 2024

 

Hola como estan? en mas de una ocacion me paso que el IA no toma los comandos negativos. al nombrar una caracteristica aunque sea NEGATIVAMENTE no la toma.
tal vez es algo que pueda solucionar

muchas gracias!

droopydog500
Community Manager
Community Manager
August 30, 2024

Hello @Camila38033095hhl4,

Thank you for your message. I am sorry you are having this problem.

 

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]", 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".  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

Adobe Community Expert (not an Adobe employee)
johnnya77883842
Participant
August 15, 2024

Would love to see negative prompting as a feature for generative AI in photoshop. Especially when using it to try and spruce up/retouching hair. A lot of times I simply type in "hair" to add more or to make hair look more tidy and it will do things like add a hair clip, or a barrette and other times it just subtracts hair.  A feature too like something specific for hair too would be nice such as changing hairstyle, adding hair extensions, etc. Thanks 

Participant
July 24, 2024

AI not understand "without" and "No" like words

 

Community Expert
July 24, 2024

Adobe is working on negative prompting. At the moment, negative words like 'no' or 'without' don't work to elmimate items. Instead, you need to describe the state without the element. For example 'man with clean shaven face' rather than 'main with no beard'.

Participant
July 18, 2024

When giving prompts like no eyewears. the image in particular generates with eyewears. That's not good.. So this would improve the image generation

Participant
July 14, 2024

If it for example generates a picture with a hat in it wich i don't want and add the text "NO HAT IN THE PICTURE" or "NO HAT OR HATS IN THE PICTURE, INGA HATTAR, HATS ARE NOT ALLOWED IN THE PICTURE, I DON'T WANT TO SE ANY HAT OR HATS" it just gives me more and more hats.

 

There should be a second text square where you can fill in what not to show or make the system understand no.

droopydog500
Community Manager
Community Manager
July 14, 2024

Hello @Martin38601577vnf5,

Thank you for your message. I am sorry you are having this problem. 

Yes, 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]", and "without [X]" does not work. As you observed, it ignores the negation word and add 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".  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

Adobe Community Expert (not an Adobe employee)
Participant
June 20, 2024

We need the negative prompts, you removed it in your new version, the advanced part, the negative prompts are an essential part of working with the AI, because a lot of times it does things you don't want it to do and that puts a stop to it.

Participant
June 27, 2024

Hi!
When are we able to use negative prompts again? Some things seem nearly impossible to generate as Firefly seems to insist on certain elements to be a certain way, with no means to alter this specific part. 
Currently I'm trying to have it generate two people from behing on a couch with one person's legs folded underneath and the other's legs are covered with an orange blanket.

No matter what I prompt, it keeps on showing feet of both of em. No feet. Show now feet. Feet are not visible...etc. etc. I even tried 'feetless' LOL. 
There are a lot of other scenarios from when I could've benefited from just one negative prompt, but it seems to be immune to all of them...