what sd version is that based on? 1.4/1.5/2/2.1?
thanks for the good work
SD 1.5 I'll add that to the model cards.
Do you have the prompt generation information for any of the sample images, particularly this one?
Do you have the prompt generation information for any of the sample images, particularly this one?
Hey, so most of the portraits shared a similar prompt structure to the following, which I kept as simple as possible to let the MJ tuned model do it's thing:
"dvMJv4, --->describe the person you want<---, 85mm, close up, portrait, telephoto"
Here's the one you quoted above:
Positive prompt:
"dvMJv4 style, lifeguard, 85mm, portrait, close up, telephoto"
Negative prompt:
blurry, soft, undeveloped, low quality, text, watermark, signature, pink, purple, mask
And here's a few others:
"dvMJv4 style, beautiful african woman, 85mm, close up, portrait, telephoto"
"dvMJv4 style, attractive future native american, 85mm, portrait, close up, telephoto"
"dvMJv4 style, mad scientist, 85mm, portrait, close up, telephoto"
Great, thanks! It was hard to identify what he's wearing, it looks like a football uniform, battle armor, or even a spacesuit. I always envisioned lifeguards wearing no shirt and bare-chested. What exactly is he wearing? Thanks for clearing it up!
Great, thanks! It was hard to identify what he's wearing, it looks like a football uniform, battle armor, or even a spacesuit. I always envisioned lifeguards wearing no shirt and bare-chested. What exactly is he wearing? Thanks for clearing it up!
Haha I have no idea honestly, I only input the simple prompt above and SD did the rest. I think it's because the images that I tuned the model on had a lot of sci-fi/random/ornate clothing, so that style carries through on the rendered images without needing to input too many specific descriptor words. To me that's part of the fun appeal of models like this is that you can get a lot of chaotic / unexpected styling.