I recognize you. It's because of your booru-forum thread that I'm even here. I've always prided myself on quality so the stricter approval here piqued my interest and will likely share here more. :p
https://guiltygear.fandom.com/wiki/Bridget lists it as a female, so if its not newhalf or futa, what is it then? Feel free to correct my tagging if you want to as i added "1girl" and "futanari" back
heart of apple 1 2, swampachine 2, sdx bulldozer 2
Ah surprised to hear more people know about swampmachine/bulldozer, tried those out a bit in the past with some mixes but never did anything further with them for now. Not sure what heart of apple is.
Are the artist tags from AID_XL supposed to work with this model? I couldn't tell, but I also don't know artists. I noticed some having some effect that someone posted on the Animagine comments as well.
No, it has no aid in it. Check the lists for 3.0 and 3.1
>>7889146 yep, also supports GPO tag, which makes all images better quality, secret tech we found out from the furry creator that he likes to hide to himself
(gpo:1.3) after the score tags om the beginning is recommended.
me neither, I just try random artist tags that I see ppl use or from animagine 3.0 and 3.1
Are the artist tags from AID_XL supposed to work with this model? I couldn't tell, but I also don't know artists. I noticed some having some effect that someone posted on the Animagine comments as well.
To anyone who want the process: 1.generate a tree 2.use controlnet canny/lineart keep all branches in same shape/place. 3.just generate 4 images with 4 seasons with that controlnet. 4.PS them together.
{Shameimaru Aya(touhou)}, full body, {raise magic}, falling down, beautiful girl, Elden ring painting style, by rolua n,by yoneyama mai, oil painting, sketch painting, {limited color sketch}, masterpiece, photo background, film still, by stu_dts,by archlich
Feel free to edit it via "Add metadata", lol. Boorus are catalogue websites, the model doesn't work if you don't tag.
{Shameimaru Aya(touhou)}, full body, {raise magic}, falling down, beautiful girl, Elden ring painting style, by rolua n,by yoneyama mai, oil painting, sketch painting, {limited color sketch}, masterpiece, photo background, film still, by stu_dts,by archlich
I had forgotten this, one of the few images generated with NAI first version where a character eats noodles properly. well, it was good while it lasted
Probably because the checkpoints are trained on danbooru tags, not the prompts you use
nah... every prompt i ever created is working like a charme on every checkpoint. "mad" is just meant i a way that those characters basically always run away or do other dumb stuff every time i prompt "getting-totally-ignored". it has nothing to do with danboorus tags
Background Creation: - Place objects in Unity to create the background and capture the screen. - Use LCM for real-time previews.
Vending Machine Photo: - Once angles are decided, take a photo of the vending machine. - Extract only the line art information from the photo, simplify it, adjust scale and placement, and add objects.
Character Movement and Video: - Determine character movements and record yourself acting in front of a smartphone camera. - Use object detection nodes to adjust proportions (similar to the method used with the snow Miku-chan). - Extract only posture information. - Use Openpose for v2v (assisted by prompts for each frame, including expression changes).
Illustration: - Composite background and character using Shoost. - Create light sources in Krita for rendering.
Retouching and Upscaling: - Retouch in Krita for final touches. - Finally, upscale the image.
Background Creation: - Place objects in Unity to create the background and capture the screen. - Use LCM for real-time previews.
Vending Machine Photo: - Once angles are decided, take a photo of the vending machine. - Extract only the line art information from the photo, simplify it, adjust scale and placement, and add objects.
Character Movement and Video: - Determine character movements and record yourself acting in front of a smartphone camera. - Use object detection nodes to adjust proportions (similar to the method used with the snow Miku-chan). - Extract only posture information. - Use Openpose for v2v (assisted by prompts for each frame, including expression changes).
Illustration: - Composite background and character using Shoost. - Create light sources in Krita for rendering.
Retouching and Upscaling: - Retouch in Krita for final touches. - Finally, upscale the image.
You are so wrong my fren. Whenever I see a girl with only animal ears visible I have to imagine she has human ones hidden by the hair, because it would be so weird otherwise. What do you imagine side of her head looks like? extra_ears is the only way to go about animal_ears.
A long time ago i found a picture of a fox lady with a ponytail on deviantart with the side of her head exposed, and it looked really nice. The author also provided a drawing of her skull that looked perfectly reasonable. Unfortunately, it dissolved in time like a steam over a kettle and couldn't find it again.
Jondor said:
I doubt even artists have any clue since they always, always cover the area with hair.
Isn't a big surprise since it's well known that drawfags are often no less lazy than prompters ๐คทโโ๏ธ Although if you commission a picture one of them with a clear request to only draw two ears, they might come up with something interesting.
if you want to know more about those prompts and my way of prompting in general, feel free to visit my thread on the forum. it's the pinned one with my name in it.
Please don't generate characters with animal + human ears, it's disgusting ๐ญ
You are so wrong my fren. Whenever I see a girl with only animal ears visible I have to imagine she has human ones hidden by the hair, because it would be so weird otherwise. What do you imagine side of her head looks like? extra_ears is the only way to go about animal_ears.
Too many toes on her right foot, and not enough fingers on her hand. Background is a little weird too... why is there a second, talley bed next to the one she's on?
Image was generated with NovelAI, edited by hand, then run through img2img in NovelAI before a final hires fix + face detailer pass to upscale and further refine with KohakuXL-delta.
after a long time, I found this, nekozoone pixai profile. this is relevant because your images here have the promps and which model was used.
this also confirms what I suspected, that his images were edited, but I never thought he was using a bad VAE. it's amazing that despite that, he manages to edit it to look decent.
for obvious reasons, I will not upload each and every one of the original images, but if anyone is interested, here is his profile. https://pixai.art/@nekozoone/artworks
ah yes, since its orange on orange its hard to see , even i myself did not catch it. right arm does indeed have wrong proportios. or rather, it has a bad render with the background. lets see if i can fix it. apologies
Wow, I am convinced this is not AI, but if it is, well done! This is really good!
It absolutely is! I spent a ton of tokens to get it all right, and then did some major edits in Photoshop, but almost every component in this image is AI generated at least~ I'm going to add a bunch of metadata manually since Photoshop wipes it apparently, so people can at least see the original before photoshop came into play for proof~
ah yes, since its orange on orange its hard to see , even i myself did not catch it. right arm does indeed have wrong proportios. or rather, it has a bad render with the background. lets see if i can fix it. apologies
