OVERWATCH - DVA -D.VA - CG like: v2.0: DvaV2.safetensors
fix the Mecha problem, more tag word makes it easier to change and easier to get those clothes off. :P
This Lora is trained based on "Overwatch Animated Short | “Shooting Star”"
The object is to make the D.va character as close to the Animated Short as possible
Newer version aims for betting fitting- expecting to be better with other Loras, Testing with Animated and realistic Loras. + no trigger word
Keeping the face and easy to change clothes to much more different styles (less overfitting), no longer stuck with the default clothing, Like the jacket or the bodysuit.
if you only want those cinematic clothes old version is better
You can still use "baseball cap" to get the hat
Don't use too many prompt about image quality or realism < those prompt is not helping, most of the checkpoint no longer need those, and most of the time it's doing more damage than helping.
ADetailer highly recommend, it is a really powerful too for fixing the face.
Using more prompts on background or clothes, default background or clothes sometimes mess up.
Working fine around 0.9, trying to get it close to the perfect weight 1 as possible, but seems like there is still a long way to go :(
Share some posts if you like bet this version can go NSFW much better :3
Hope you like it, I know I'm making too many versions :( (still learning)
Join the discord if you like, Some time I post my newer Lora and image there for test :)
I always wonder why I can't get the facial mark which part of d.vas famous styles, and I find out she doesn't even have the mark when in cinematic :<
-
-
-
-
-
-
-
Old versions --------------------
I made a discord server, it's in my profile, if you want to chat, share some ideas, share some images, or if you want me to do some Lora, feel free to join.
v5 - changing the image set, less mecha, overall improvement, training is different, should be able to make the lora more flexible, background change, should be much more easy, changing the trigger word to only owdva, should be less mixing with the 1.5 trigger word.
Adding long hair back so if you want the "baseball cap" use the trigger word, or it will be random.
And, NSFW support is much better :^ So I think clouth should also be much easy changeable.
share some posts and join the discord server :^)
Sooooooooooooooooooooooooooooooooo
v3- is overall much much better than what I expected, It should fit all 3D Models, with no lower-level body, which means no bodysuit, and no long hair picture, so that baseball cap is stuck on there and no longer needs a trigger word. Think this should be the last version on Dva- CG-like, those 3 versions should have all the options.
V2- as I say in V1, I do menage to take off D.va's Mecha from the Lora. retake all the training pictures, and tag them out one by one.
weight between 0.45 - 0.7 should give a great result.
Tag out those hats, and suits in the training, if you want to show it, put the trigger word in
pro: No mecha, more pose, easier to change those clothes.
con: I think is less CG-like, background is mostly black and poor finger (but you can use the glove to cover it), For some reason she keep doing that (WTF pose) I will go see if I can take those off, but less the lora weight should less the chance of it.
Share a post if you like them :3
suggestion is always welcome :D