Type | |
Stats | 33 6 4 |
Reviews | (4) |
Published | May 8, 2025 |
Base Model | |
Training | Steps: 910 Epochs: 1 |
Usage Tips | Clip Skip: 2 Strength: 0.9 |
Hash | AutoV2 C95C2AEE1D |
Let's make SD1.5 great again!
V1.2 just released, trained with captions this time and on almost the same image set as V1. Only 910 steps so it shouldn't be over trained. My sample images is generated in 768x1024 and no post-work whatsoever. Most SD1.5 checkpoints can't handle that large (large for SD1.5) image size so keep that in mind.
Image got shrunk so not so easy to see anything but there are a lot more coral and fishes and more shading on the images with higher strength.
V1.0
I experimented with a LoRA that should add more juice to images, seems to work.
V1 is without captions so it can sometimes mixup stuff like geckos with dinosaurs.
No post work like aDetailer or inpaint on any of the samples images or examples below.
Example images below, left side without V1, right side is with V1 and exactly the same prompt and guidance: