Here some ideas to make Adobe Firefly Better

[BangWee]

thank you for inviting me to be a beta tester for adobe firefly

I explored the Generated Image with AI website a lot. I even use stable diffusion to make really good pictures. either realistic photos or vectors.

some of the shortcomings that I see from text to image
1. It seems that Adobe Firefly has very little material, for example: if I use the command "Create a realistic image of a young asian woman sitting on a park bench in autumn. Use a Hasselblad camera with a 85mm lens at F 1.2 aperture setting to blur the background and isolate the subject"

the face of an asian young woman, must be the same even though it generates a lot of times.

there may be a need for a lot of resources and of course a lot of users so that more and more materials.

2. There are some things that I found that Adobe Firefly can't make.
- chibi (in generated many times it looks the same)
- logos

3. Adobe Firefly doesn't have a detailed description of prompts, such as midjourney https://docs.midjourney.com/docs/prompts

4. Please add a negative prompt, sometimes if you want to make a human, sometimes a human has 3 eyes, 4 eyes, 3 cheeks, 4 hands, 3 legs.

5. detailed explanation of Aspect ratio, Content type, Color and tone, styles, Lighting and Composition

6. I hope there is a video to img generated

7. add stock costumes, or any object. if you've ever played the sims, like changing clothes, I think if that feature exists it can be used with any object, be it plants, leaves, humans, objects or whatever so without any prompts we can enter anything we want.8. I'm still waiting for all the features in Adobe Firefly.

[BangWee]

and i have 1 more ideas,

is consistency image generated  

[BangWee]

[BangWee]

Ohh okay i got it,, Adobe Firefly have featured use reference image to make consistent image.
sorry im new member here, i dont know how to edit the post ahahaha thank you 

[Tracy.H]

Thank you for taking the time to provide feedback! I have a couple responses for you:

1. At this moment, we produce art better than photo realistic images. We are working to improve this.
2. I will suggest "chibi" as an art style suggestion. For logos, we are exploring text to vector which would be better for logos.
3. We are considering some tooltips for prompting.
4. I don't think negative prompts will necessarily improve the number of limbs or eyes on a person. This is an example of needing more training and improving our model, which will come with time.
5. Tooltips for different styles and characterstics is a good suggestion.
6. I'm not sure how video to image would work? If you meant text to video, this is highly requested feature.
7. Stock costumes is a new idea I haven't heard yet. We can consider this.