hey looks like you forgot to add prompts to your starter images for people to test the model with. What kind of dataset did you train this on?

I have tested around 12 pony checkpoints in the last five weeks, and I did it thoroughly! The main goal was to find a checkpoint that lived up to its own claims in terms of being 'realistic'. Well, we all know that a pony checkpoint has to make a lot of compromises before it is born and so it is always a balancing act as to which of the desired elements gets 'more weight' in the end. We also know that you can never please EVERYONE and so it is up to the creator to ultimately decide for themselves. At the beginning of my 'search' I was naive enough to believe that the first checkpoint that drew the most beautiful and perfect hands during my first generations was now standard and I was content with that until I continued my curiosity and looked a little more closely at similar results here on CivitAI. From that moment on, not a day went by without me comparing all my renders with unknown 'checkpoints' because I am a 'researcher' by nature. This led me to simply raise my expectations to a realistic level and I began to separate the wheat from the chaff. I also know that hardly a day goes by without new checkpoints being created, old ones getting an update, etc. but from my current perspective I can say with certainty that realisticPonyPhoto V10 is by far the most 'realistic' checkpoint that keeps its promises! There are a few minor errors related to the prompt design (e.g. several people in a landscape format, the eyes that are then immature and have different colors, drooping eyelids), but the responsibility is simply amazing! In my case, I found that 'wet' and 'oily skin' lead to an overall 'cleaner' result and give the image a positive touch, whereas without these prompts it looks rather 'bare'. As a result, I am already curious if there will be a sequel to this checkpoint and would like to stress that if there is, I will gladly invite the creator for a drink of his choice! Cheers!

Very impressive; the other Pony i used took forever to render, this one is on par with 1.5, but with crisp results. What´s more important is that it actually listens to prompting. Thank you, i can now delete 20 GB of useless models.

Great model, really good prompt accuracy but what stands out in the generalisation ability. It is not heavily biased and i can use it to change details in images in inpainting and use photoshop to merge so much better than with every other custom model I have tried.
And its fast on my old 2060 12gb.

Love the model ❤️.

So far it's my favorite for any anime/game -> realism visualization, and from just tinkering around, it is a great model all-round.

Thank you so much making and sharing it!