Wow, this was really awesome in results! @ausbitbank might be interested in this model! Excellent work!
You are viewing a single comment's thread from:
Wow, this was really awesome in results! @ausbitbank might be interested in this model! Excellent work!
:D I love that despite how much effort and time it took, it is something that you like. I hope to make more good images :)
I appreciate your appreciation.
Next time, if you want, add those smaller ones into your post as well. These did all turn out pretty amazing! Thanks for sharing! Looking forward to seeing more of your outputs. I'm sure you probably already know... but if you find images you really like and have access to the seed... You can keep up with and reuse the seed so you don't have too much of a variance on the next run.
I still don't know much about the influence of seeds. I do reuse seeds for the same prompt or variations of it, but for completely new prompts I've never tried reusing an old seed because I don't know if there will be any advantage to it. I tend to just leave things to work 20-60 times and grab the best results and make bigger and more detailed derivatives, I go for 15-30 of those and then post the best ones.
Do you know more about what the seeds do if the prompt is completely different?
The seed will keep a similar style to what you have previously seen as the results. So, if you do a completely different prompt, then you might get some similar primary features or a similar style, even though the prompt is different. You will have to play around with the prompts with a same seed. Like do three different prompts with a seed that has produced results you like prior to running your test runs.
Seeds, to me, are the biggest mystery of this whole process. I have no idea where the seeds come from or why using the same one creates some continuity in a baseline... but once you know, the same seed creates similar results... then whe you find good results, perhaps save that seed and try it out again on completely different prompts and see what you can come up with!