Firefly is becoming impossible to use with every passing day.

[ravi_2912]

This prompt is not generating result - " Modern Japanese animation style illustration of a determined boy steering a small wooden boat through stormy waters. The boy should have detailed black hair blowing in the wind and a focused expression. Clean linework with vibrant colors, featuring a character with slightly exaggerated proportions, large expressive eyes, and simplified yet emotional facial features. The background should be turbulent ocean ONLY with stylized blue waves and white caps. Dark storm clouds above with dramatic lightning. Rich color palette with strong contrast between dark waters and highlighted areas, attention to light and shadow detail. Cinematic composition with dynamic movement lines and the emotional storytelling aesthetic typical of contemporary Japanese feature animation. " 

firefly image model 4 can't generate it, model 3 generates but of poorer quality. Till Few days back model 4 was working fine for me. What's the issue, is there something problematic in model4 that it cant generate the image, whilst  model3 can generate.  

[Oh.N8]

Hey @ravi_2912 

Thanks for sharing your prompt, this can be frustrating for sure. 
I find using simpler prompts works well. 

Start with a base image and build on the image by using various images as style reference and composition reference.

Cheers

O. Nate 

[droopydog500]

Hello @ravi_2912,

Thank you for your message. The examination of prompts for compliance with the gen AI guidelines in Model 4 is clearly changed from Model 3. I have to say this is one of the worst cases of inappropriate blocking I have seen. It took me over 50 generations to narrow this down. Note, I did flag every single failed generation:  

In the end, these are the words that triggered the error:

Modern Japanese animation style illustration of a determined boy steering a small wooden boat through stormy waters. The boy should have detailed black hair blowing in the wind and a focused expression. Clean linework with vibrant colors, featuring a character with slightly exaggerated proportions, large expressive eyes, and simplified yet emotional facial features. The background should be turbulent ocean ONLY with stylized blue waves and white caps. Dark storm clouds above with dramatic lightning. Rich color palette with strong contrast between dark waters and highlighted areas, attention to light and shadow detail. Cinematic composition with dynamic movement lines and the emotional storytelling aesthetic typical of contemporary Japanese feature animation.

I was finally able to generate with this prompt:

Modern Japanese animation style illustration of a determined boy steering a small wooden boat in a storm. The boy should have detailed black hair blowing in the wind and a focused expression. Clean linework with vibrant colors, featuring a character with slightly exaggerated proportions, large expressive eyes, and simplified yet emotional facial features. The background should be rough sea ONLY with stylized blue waves and white caps. Dark storm clouds above with dramatic lightning. Rich color palette with strong contrast between dark seas highlighted areas, attention to light and shadow detail. Cinematic composition with dynamic movement lines and the emotional storytelling aesthetic typical of contemporary Japanese feature animation.

My apologies for this over aggressive blocking. I will pass your original prompt over for further examination as how this aggresive blocking can be tamed.

My best,
    droopy

[ravi_2912]

Thanks droopy, I had used gpt to generate the prompt, once i gave it my visual ideas. I did tweak the prompt a lot but was still not able to generate results.

[rhodie62]

What a huge dissapointment. No more quirky or unusual imagary with mysterious and whimsical backgrounds. The results in image 3 are utterly unexciting and predictable and worse the figures and faces are often contorted and the faces are often distorted. I thought image 3 was a step up on 2 but thats not the case. The results are bland and very far from artistic and imaginative. Worst of all I continually get "can't load" messages for reasons I can't fugure out. This message seldom came up when using image 2 because I knew how to promt without generating this message. Please bring back image 2. Image 3 and 4 are not an improvement - they are an imaginative creators deathknell - dull and insipid. Why can it not beprovided side by side as an alternative rather than scrapped.

[Kartika Rawat]

Hi rhodie62!

Thank you for reaching out and sharing your feedback about Image 3. Would you be willing to share some images and the prompt with us? Regarding your suggestion to provide Image 2 alongside Image3 and 4, we will certainly pass this on to our development team for consideration. 

In the meantime, if you continue to experience issues or have specific prompts that are causing problems, please let us know. We are here to assist you and ensure you can continue creating with ease and satisfaction.

KR

[rhodie62]

Hi Kartika

Thanks very much for getting back to me - I really appreciate your interest.

I have attached 1 image from a previous image 2 generation that I made a few weeks ago plus I have attached the 4 images generated on image 3 all from the same generation prompt. The prompt is as follows:-

Photo of an unhappy unsmiling very sad teenage Caucasian girl facing the viewer dressed in traditional scandanavian dress with a bonnet on her head sitting on a low short stool squarely facing the viewer directly her feet in boots on the ground with soft studio lighting in a old darkened room with old soviet military furniture and crumbling and cracked cement walls with flakey peeling paint with an open balconey archway in the back wall overlooking a bullfighter in the ring behind the girl blue sky

 I have obviously used words that may seem unusual but after using image 2 for more than a year I quickly figured out how to get what I wanted with those words - which for the most part I use when generating images of people.

 As you can see in the image 2 image there is a beautiful harmony of colour and format and proportion as well as an interesting and textured back ground that is very evocative with a sense of place and nostalgia. These are all elements that I as an artist seek to find. When you look at the image 3 images they are utterly devoid of interest with distortions as if they were just dropped into a background with no thought of a sense of place. There are many subtilties in the image 2 image that are very pleasing and ultimately harmonious to the whole image. Subtilties that are definitely missing from the Image 3 images. The prompts I used in this instance are - Baroque, Surrealism, paper mache, Industrial, Dark, chiaroscuro, Daguerreotype and photo.

 This is one of many thousands of images I have created that share the same unique appearance.

 Image 2 is an incredible tool for artists who do not just seek a photographic image as an end result. Image 2 gives the artist so much more incredible information in terms of setting and expression and harmony and proportion and a sense of mystery and much more besides as well as quirky and whimsical distortions that add a sense of mystery and charm to the overall image. Not having access to image 2 is definitely going to impact my work and I am already looking at other AI tools - something I never needed to do because image 2 gave me exactly what I wanted and needed. I doubt also whether other AI tools will give me the same incredible nuances that image 2 gives.

 I do not copy an image directly from prompts but rather generate many prompts from which I take bits and pieces along with other imagery I create to make a final painting. I swap heads and hands or backgrounds and skies on PS until I have the final image which I then use as a master for a painting.

 This is a fantastic tool and one that as I read through the forum I see many artists are going to miss. PLEASE, please bring it back. It is so much better than 3 and 4. This tool should be in addition to 3 and 4 because it is so unique. It will be wasted and deeply missed.

Thanks for taking an interest in my complaint (or rather compliment because it is such an amazing tool that PS has created.)

 Sincerely

 Allen Egan

(Artist)

[Kartika Rawat]

Thanks, Allen Egan, for the screenshots. It helps. We really appreciate your passion and your kind words about Firefly. Your feedback has been shared with our team, and we’re actively exploring ways to better support moving forward.

KR

[rbb_8750]

I really don't know what's going wrong here, but I'm slowly losing interest in Firefly, I've been working on it intensively for 2 days and I've never had so much stress with any Ki.

Why can't I generate a "colored boy" for a children's book? What's wrong with this prompt?
“A little black boy with short curly hair, brown skin, a red T-shirt, shorts and sandals stands in front of a mud house in a sunny village - in the style of a children's book, friendly, soft lines.”

permanently, at whatever comes the message :

Unable to load
We are unable to process this prompt. You can edit it and try again, or mark it for review. Note that prompt data is collected when it is marked for review.

This is also not possible:
Prompt:
Watercolor drawing of a little boy with dark skin and curly hair. He is sitting under a tree and drawing on the ground with colored pencils. In the background is a simple mud house, warm colors, friendly style.

My system is top, my internet is top? What's the problem?

And now this prompt also not works:

Illustration of a snowy street , a few old and dirty apartment building with rooftops covered in snow, bare trees, children’s book style, soft cold colors.

How i can get my money back, this software dont work! I loosing time and my nervs. i dont want pay for this 219,98 €. 

[SamarthS]

Hi @rbb_8750!

Thanks for reaching out and sharing the prompts. I tried the prompts and got errors using the Model 4, however I was able to generate the images using Model 3. The screenshots of the images are shown below. Please try using Model 3. 

The product team is already aware of the issue. Meanwhile, please use model 3 to generate images 

I am merging the thread with similar discussions.

Thanks

Merged thread