Exit
  • Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
  • 한국 커뮤니티
2

Firefly needs to drastically improve its training data.

Explorer ,
Apr 06, 2025 Apr 06, 2025

I have never had any good result with Firefly, no matter how detailed the prompt is or how simple the prompt is.

 

1. Firefly cant spell text correctly, okay fine, but other AI apps do. 

2. Firefly generates inconsistent results  when it comes to demographics, sometimes the results will be fine, sometimes it will be straight out of the park. E.g. 

2a. Pretty okay results.

ravi_2912_0-1743968551947.png

2b. unbelievable results , how incredibly bad.

ravi_2912_1-1743968654650.png

Just see this and guess is it some Alien image ? 

ravi_2912_2-1743968725393.png

The faces are rendered horribly and the attire in all the pictures are totally wrong. 

 

I am surprised Adobe has so much commercial resources but still  Firefly is so lacking, is it lack of interest in AI product development or some other barrier preventing Adobe. 

 

Idea No status
TOPICS
Imaging
48
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

Adobe Employee , Apr 07, 2025 Apr 07, 2025

Hi ravi_2912!

We appreciate you writing in. There are some limitations in the generated results. Distorted text in generated images: We recognize some irregularities in certain scenarios, but text and symbol generation in images still needs support in the Text to Image feature. You can report images with distorted or unclear text using the Report icon, which appears when you hover over a generated image. You can import images in Adobe Express and text. Unrealistic body features in generated imag

...
Translate
1 Comment
Adobe Employee ,
Apr 07, 2025 Apr 07, 2025
LATEST

Hi ravi_2912!

We appreciate you writing in. There are some limitations in the generated results. Distorted text in generated images: We recognize some irregularities in certain scenarios, but text and symbol generation in images still needs support in the Text to Image feature. You can report images with distorted or unclear text using the Report icon, which appears when you hover over a generated image. You can import images in Adobe Express and text. Unrealistic body features in generated images: The quality of our models will improve with time, and generated content will continue to better follow prompts. Select the Report icon to provide feedback about distorted facial or body features.

 

Let us know if you have any questions.

Thanks,
KR

Translate
Report