Trying out dynamic prompt templates. Seems fun, and it's an easy way to produce tons of fun images, but cherry-picking becomes extra long and tedious process. 😆
Damn, I need to finish playing the trilogy. 😆
Included files: 1-6. Young Spyro 7. Prompts metadata
Gah Damn...I have so much Spyro pictures up my Sleeve, I really cant count it anymore.. XD Should have posted at least one of em.. XD He looks really Great, Autosnep! ❤️
Gah Damn...I have so much Spyro pictures up my Sleeve, I really cant count it anymore.. XD Should h
My Version is a lil bit ... Taller then yours.. tbh. XDD Will Upload him maybe tomorrow If I find Time. But Yeah... his Fins... Wings and such.. are really a big part to edit.. XD
My Version is a lil bit ... Taller then yours.. tbh. XDD Will Upload him maybe tomorrow If I find Ti
Serious question here. I've noticed that in a lot of AI generated sequences, it's just the same image with slight tweeks and color / lighting / shading differences. How come no over head shots, worms eye view shot or rear shots. Are they too hard to do in the same sequence or what. The images themselves look great but how about some variation. :)
Serious question here. I've noticed that in a lot of AI generated sequences, it's just the same ima
Humans tends to prefer front-facing shots for virtually everything, on average. 😁 They look the best to us (humans love looking at faces), so usually this is what images and photos get posted online, what images are chosen for training datasets, and thus, what Stable Diffusion models understand the best. 🤓 In general, Stable Diffusion generates "typical" images at highest quality: front view, simple pose, no obscured parts etc.
So, when what people want the most (frontal shots) and what models can generate the best (frontal shots) happens to match, what you get is, well, frontal shots. 😁
It isn't impossible to generate rear shots, low-angle shots and whatnot, and in fact I have several images with just that ready to be posted, but it's harder to get them right and they're likely to attract less attention.
By the way, the same applies to pre-AI artists as well. A lot of people stick to simpler camera angles because it's much easier. Current AI models can't handle generating proper perspective because they can't draw perspective lines, but let's be real, vast majority of living artists never learned to do it, and if they did, they're too lazy to bother with it. 😆
Humans tends to prefer front-facing shots for virtually everything, on average. 😁 They look the best
Makes sense. You do what the customer wants. As for perspective I never had that much of a problem with it once I learned the rules. I'm kind of an anomaly in that I can see perspective (when I need to) on simple views without needing to grid it off. But ya perspective is fun / hard for some and I can totally see needing to train AI more so it understands what it's putting down and why. :)
Makes sense. You do what the customer wants. As for perspective I never had that much of a problem w
What's funny about Stable Diffusion is that it copies how human artists draw things.
If you ask for a butt focus shot, it'll balloon the butt out of physical proportions. If you ask for a big butt, it'll bend the character over and break its spine to make the butt stand out. If you ask for pussy and anus in a complex image, it'll happily move genitals around the crotch to make them face the camera.
Why? Because artists do this. 🤣
So even if you do have different standards, you have to overcome the model's habits. 😆
What's funny about Stable Diffusion is that it copies how human artists draw things. If you ask for
