Verified: 4 months ago
SafeTensor
Type | |
Stats | 297 0 |
Reviews | (15) |
Published | Jan 22, 2025 |
Base Model | |
Training | Steps: 600 Epochs: 200 |
Usage Tips | Clip Skip: 1 |
Hash | AutoV2 B94F8AF975 |
I am releasing an unplanned realistic model. I have not had a release for some time. This model is unique in that it was fully fine tuned on hand picked images and data at 768x1024 and 1024x768. This allows this SD 1.5 model to make up to 1024 images without high res fix. The preview images do not have any LORA, embeddings, or anything on them. They are only highres fixed at 1.5x (1152x1536) from the previous stated base resolutions. The only exception is the long distance night galaxy image was inpainted to fix the details, one pass for the full model, and one for the face. I am honestly impressed with the detailing in some of the images this creates.
Prompts can be simple to advanced. No VAE suggested or required.