Giantess
A LoRA model for generating giantess.
anima-v1(v0)
This version is only compatible with anima-related models. Prompts should be written using a combination of tags and natural language descriptions. This version has a very high composition ability (compared to the Illustrious version). You can use natural language to describe the size of the giantess, expressing it through her relative scale to objects like:
she is as large as a building
she is larger than skyscrapers
tiny people as large as her breast size
she is larger than a city
mountains are visible below her
planet is as large as her feet
Only using tags or too few prompt words may lead to poor results.
Suggested Generation settings:
CFG: 4~6
Strength: 0.8~1.0
sampler: er_sde
steps: 40~50
v2.0 (Illustrious)
An enhanced dataset for city-size giantess. For giga size / planet size / shrunk size, v2.0 does not guarantee better results than v1.0.
This version improved the variety and density of urban landscapes, including details like skyscrapers, tokyo tower, bridges, and highways. The airplanes and helicopters are now much more stable. Additionally, the visual rendering for dirty feet and smell effects has been intensified. The destruction effects and interaction between buildings are improved.
V2.0 Recommended weights: 0.7 ~ 1.0
Add more details in your description. Insufficient prompts may generate really bad images.
Add details to your prompts instead of just typing the trigger words. For example, you can generate city-sized giantess using 'giantess, city, building, skyscraper, tokyo tower, highway, bridge, destruction', etc., without specific trigger words.
giantess: General words, used for all images.
giga size
mega size
building size
planet size
shrunk size: Normal size girl interacts with shrunken city/people.
V1.0 Recommended weights: < 0.9
Description
An experimental version, really unstable.
FAQ
Comments (14)
:( pleeeease can u do it on stable 1.5.... please...
你把我哪些胡乱想到的实现了一部分，非常感谢
can you do a regular lora on sd 1.5 please
I can't get this lora to be recognized by automatic1111. How can I fix this?
You need to have an SDXL-type model loaded, not an SD-1.x model.
why i can't find this lora from resource?
Since I've seen some images in the gallery using Pony XL V6 based checkpoints: This is NOT a Pony based LoRA so it will have no (positive) effect on your images with Pony. Pony (and Autism) already know the concept so it will work without the LoRA. Use this LoRA with Kohaku-XL.
a lot of xl loras will work as intended with pony, probably because pony is xl based in the first place. not true for all, but you can check by running gens of same parameters with and without lora.
@lilquin I tested quite a few initially and none of them worked for me as intended by the LoRA creators. There is a reason why Civitai decided to list Pony as a base model.
I mean, it says it right there that it is an illustrious model.
Also, afaik pony only knows non-extreme sizes (building/doll) so it's useless if you're seeking extreme sizes
@ididididi It's a 9 month old post, Illustrious didn't even exist back then and again: Pony knows the giantess concept because it's a well known tag in danbooru which was part of the training dataset for Pony XL V6. A good pony specific giantess LoRA would improve the generations though, depending on what you are aiming for.
@dude_ Ahhh gotcha, sorry.
Also yeah, its a pretty popular tag, but only in general.
Extreme size difference for example is a frequently misused tag (many, MANY people flood it with normal, building sized giantesses) so its a no way without a lora.
Also, there arent any common tags for specific scenarios (and sadly, no Lora for any of those either. I hope to change that but it's been quite challenging so tar lol)