Updated: Nov 9, 2023
conceptVerified: 2 years ago
SafeTensor
Type | |
Stats | 523 1.5k 8 |
Reviews | (59) |
Published | Nov 7, 2023 |
Base Model | |
Training | Steps: 12,810 Epochs: 210 |
Usage Tips | Clip Skip: 2 Strength: 0.8 |
Trigger Words | titleConcept,oversized shark behind girl,evil shark behind girl,standing,looking at viewer, rideShark,riding on shark,straddling on shark,underwater,straddling,riding,dark water,dark ocean,oversized animal,size difference, fightShark,girl fighting a shark head-on,girl glaring at a shark,from side, EatPerson,girl in shark mouth,shark opening mouth,girl in big shark mouth,space,iceberg,air bubble,dark water,dark ocean,oversized animal,size difference, MeetShark,girl in the back glaring at a shark,girl meets a shark underwater, from behind, looking away, |
Hash | AutoV2 DDB09B220A |
LoRA (36.12 MB)
Verified: 2 years ago
SafeTensor
I created a combat scene with a shark using LoRA. I aimed to display the following five concepts:
1, A pattern where the LoRA character is inside the shark's mouth.
2, A composition that resembles a movie title, using the LoRA character and the shark.
3, A pattern where the LoRA character encounters a giant shark.
4, A pattern where the LoRA character is riding on top of the shark.
5, A pattern where the LoRA character and a giant shark start fighting side by side.
For the scene where the character is about to be eaten by the shark, a vertical image format tends to produce images with the character inside the shark's mouth, whereas a horizontal format tends to warp the shark's mouth. Additionally, during the operation check of LoRA, I noticed that using a size of 512x768 often resulted in the LoRA character's face being distorted. Therefore, a size of 1024x1024 tends to yield cleaner images.
Show more
