• 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
0

Firefly disregards prompt completely, if there is some reference and style images added.

Explorer ,
Jan 23, 2025 Jan 23, 2025

Copy link to clipboard

Copied

I never found Adobe Firefly to be of any use, I saw many designers being fearful of it replacing humans and all that. 

I tested it in all permutations & combinations, but it was not able to generate images of even very very simple prompts of simple objects like a pastry.  See my Screenshots below.  

ravi_2912_0-1737704919789.png

ravi_2912_1-1737705048751.png

 

None of the pastry looks purple in color.

After every vague result, I shared the feedback with adobe too. But frankly I would rate firefly to be totally disregarding the user prompt.

TOPICS
Imaging

Views

67

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
Community Expert ,
Jan 24, 2025 Jan 24, 2025

Copy link to clipboard

Copied

Hey @ravi_2912 

The images that you are providing are red. Use these a Composition ref 100%. I find using a colour palatte in your Style Reference helps. 

Prompt
cake with layers that are green, orange, red

Screenshot 2025-01-24 at 09.24.08.png

Screenshot 2025-01-21 at 20.41.01.png

Screenshot 2025-01-24 at 09.26.14.png

 

Cheers

O.Nate

Votes

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
Explorer ,
Jan 24, 2025 Jan 24, 2025

Copy link to clipboard

Copied

LATEST

But, Why can't firefly read the prompt the user gives ? If it can't process the prompts then why it asks for it, it should clearly state that it will generate images as per some variations  to reference image provided. ( Even Google Image Search has that feature since many years). 

Firefly model is trying to straddle 3 user inputs - reference image, style, and prompt and the result comes out totally indifferent to what user wants. 

Firefly is highly disappointing and no amount of justification helps. I sometimes get better image result from Co-pilot. Not at all in league with midjourney , kling.  That's my most honest feedback.  

Votes

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