Type | |
Stats | 1,699 528 |
Reviews | (200) |
Published | Apr 18, 2023 |
Base Model | |
Training | Steps: 4,000 |
Trigger Words | T800Endoskeleton robot with a metal skull face chrome metal body holding a gun holding a futuristic gun shooting lasers |
Hash | AutoV2 9BDCFCA0D3 |
After scouring through hundreds of models and trying to generate a Terminator Endoskeleton it became obvious to me that very little of Terminator exists in any model's training data. So, using Hollowstraberry's tutorial & colab I trained this LORA on about 80 HD images at about 4000 steps on Experience 8.0. It does closeups the best, but full body shots are still quite good, and seems quite flexible with styles and different models.
I recommend a weight of 0.8 - 0.9
Adding "holding a gun" or "holding a futuristic gun shooting lasers" without quotations should produce the signature futuristic rifle the Terminators usually use in the future.
I recommend 512x512 or 768, LORA was trained at 768 resolution so does well with higher resolution. It seems to handle high res fix really well and produces amazing quality with it.
This is my second LORA and I don't really know what I'm doing outside of the tutorial and colab, I have a AMD 7900xt so I have to use the DirectML version of A1111 whichi doesn't currently work for local training for me. If there's any tips you might have, please let me know.
Please rate & comment and share your generations & let me know how I'm doing!