licking a toilet bowl版本ver0.6 (ID: 1104629)

It has long been said that the discovery of a new picture of a girl licking a toilet bowl is as valuable as the discovery of a new celestial body. Now we have AI and we have LoRA. Let's create the sparkle of stars.

Dear all. It is a pleasure to see you again. I have made an update in search of a better LoRA. I have made a radical revision of the dataset and have succeeded in improving some features. However, there are still some shortcomings that remain. In particular, including more than one person in an image sometimes produces a bad anatomical picture. For this reason, I am marking this version as 0.8.

recommended weight: 1

Trigger word (example)
licking toilet, toilet, all fours, looking down, open mouth, tongue out
licking urinal, urinal, wariza, looking down, open mouth, tongue out
licking stall, stall, all fours, looking down, open mouth, tongue out
licking floor, stall, all fours, top-down bottom-up, head on ground, looking down, open mouth, tongue out

How to use this LoRA and what updates are available

Dataset improvements have reduced the influence on the art style. It is now less likely to output photorealistic figures despite the use of cartoon-like checkpoints. However, it is still not perfect. This is due to the fact that there are few anime-style drawings with accurate toilet bowl structures and the dataset is biased toward photographs. I would like to continue to improve it.

Able to choose the type of toilet bowl.

In Ver0.6, there were two types: sitting type and standing type; in Ver0.82, there are now three types. Let me introduce them along with their trigger words.
toilet: A type of toilet that is used sitting down.
urinal: A urinal hung on the wall
stall: A urinal placed on the floor
Now let's write a prompt for licking.
After selecting the type of toilet bowl, add the trigger word.
toilet, licking toilet /urinal, licking urinal /stall, licking stall
Please prompt for the appropriate posture for licking the toilet bowl. Such as "all fours", "wariza", or "on stomach". I also recommend adding instructions to tilt the upper body, such as "bent over", "head down", "looking down", "open mouth, tongue out" etc.
Camera position is also important: "from side", "from behind", "from above" or "from below".
With the update, you can now use “facing away”. If you use “from behind” and the person still looks at the camera, try “facing away”.
You can now use “grabbing another's head” and “stomping another's head”. I recommend using “head out of frame, male head out of frame” in combination. The use of “silhouette man” may be better, since it is difficult to indicate costumes for the hands and feet.

Of course, “licking floor” is also possible. Use “top-down bottom-up, head on ground” together. And It's time to use “stomping another's head”.

Below is the description for ver 0.6

Known Problems
All images in this LoRA dataset are photos. This is due to the fact that I have found very few animated images depicting a toilet bowl from below. This caused the background to be extremely photorealistic. This is not a big problem with photorealistic checkpoint output and characters that illustrious can output with tags, but when combined with LoRA for ANIME characters, the colors are anomalous.
You will need to lower the LoRA weight and generate a lot of them until you get a hit.

Recommended weight 0.8-1

Trigger Tags
licking toilet, toilet
licking toilet, licking urinal, urinal
licking toilet, licking floor, urinal

Tags that should be added
all fours, open mouth, tongue out, looking down, head down
(Instruct the appropriate posture for licking the toilet bowl.)

option tags
grabbing another's head, sticking head

Translated with DeepL.com (free version)

描述:

Released as ver0.6 due to a problem when combined with some LoRAs.

训练词语: licking toilet, toilet,licking toilet, licking urinal, urinal,licking toilet, licking floor, urinal,grabbing another's head, sticking head