Current state:
Using the gradio text to image (i think img 2 img will be the same) the programm creates a folder for each prompt. In order of doing so spaces will get replaced with _ (and possibly even more once #94 is fixed). This results in prompts with havily different outputs ont the same seed to be identical (e.g. A ice phoenix, epic, fantasy, 4K, render, cinematic vs A_ice_phoenix,_epic,_fantasy,_4K,_render,_cinematic both save into the same folder with no distinction, but here are examples on seed 274725
)
Possible solution
Add some exif or itxt data to each image which includes the original prompt and possibly also the seed.
other
I am currently working on implementing this change and will create a pr soon
Current state:
Using the gradio text to image (i think img 2 img will be the same) the programm creates a folder for each prompt. In order of doing so spaces will get replaced with _ (and possibly even more once #94 is fixed). This results in prompts with havily different outputs ont the same seed to be identical (e.g.
A ice phoenix, epic, fantasy, 4K, render, cinematicvsA_ice_phoenix,_epic,_fantasy,_4K,_render,_cinematicboth save into the same folder with no distinction, but here are examples on seed 274725)
Possible solution
Add some exif or itxt data to each image which includes the original prompt and possibly also the seed.
other
I am currently working on implementing this change and will create a pr soon