PREMATURE, TAB PROBLEM, misspellings and not using provided photos for modeling

[monty257723240eb1]

3 times now when I am typing a prompt the engine begins rendering simply because I pressed RETURN to TAB down in the wring of my prompt.  I do not feel I should be charged for this.  Maybe have a specific KEY to start the RENDERING.  

When I provide a photo I expect the engine to use the photo, NOT be inspired by it.  This work for a project and I cannot use aproximations or Inaccuracies.  Seriously, there MISPELLINGS in areas where I provide the exact spelling and layout and photo.  

What cinematic type animation can I use that creates with all of the important factors included correctly.  I am talking about about historical aircraft and not some fantasty world.  

I also should not have to pay when I have not said GO.  There is no way to stop.  No :05 delay, nothing; I would like to be credited. Why is it that every other type software I work with will ask me three times if I am sure that I want to do what I just said I wanted to do. 

Thank you,

Monty

[MicahBurke · Accepted solution]

A lot of this has to do with the model you're using (Firefly vs ChatGPT vs Gemini).  Not every model excels at generating images perfectly, nor good at text.  Each model has it's own strengths and weaknesses.

For text, I recommend GPT Image or Gemini 2.5. As you can see it does a good job when given the prompt:  "make circular stylized text with a 20-sided die in the middle, "is it my turn again?""

Firefly 4 got the text ok, but the rest of the image is useless.

Gemini 2.5 (Nano Banana) made an awesome image, but not really what I was after (was wanting more of a stylized vector-like image, had I prompted for it, Gemini probably would nail it.). It also got the text spot on.

Flux Kontext... well it's not the best at this kind of thing:

This guide can help you decide which is the best model for you:  https://www.adobe.com/products/firefly/partner-models.html

I agree with the tab/enter button and with the payment issue.