Verified: 2 years ago
SafeTensor
Type | |
Stats | 3,543 55,776 127 |
Reviews | |
Published | Sep 10, 2023 |
Base Model | |
Training | Steps: 4,880 Epochs: 5 |
Trigger Words | underwater submerged |
Hash | AutoV2 F48246C1AD |
This LoRa is trained to produce underwater pictures. It was trained on 400+ manually selected and cropped pictures. It works best with 512x768, 768x512 and 768x768 resolution.
Note : if you want to try to create images in a swimming pool, do not use the tag "swimming pool", use "floor tile" or "tile" instead (and/or "walls"). This is because your Checkpoint most likely associates "swimming pool" with surface pictures.