Type | |
Stats | 402 |
Reviews | (56) |
Published | May 27, 2023 |
Base Model | |
Training | Steps: 120 |
Usage Tips | Clip Skip: 1 |
Trigger Words | dgstyle |
Hash | AutoV2 2F297C0882 |
There is a criminal conspiracy embedded deep in the law enforcement, military, and intelligence services of the United States. An institution dedicated to facing threats so dangerous that mere knowledge of their existence is cancerous to the human mind. Illegal, deniable, dedicated and decisive, the men and women of Delta Green put their lives and sanity on the line to keep the clock from striking midnight.
For today, at least...
This embedding is trained on the works of Denis Detweiler, a primary author and artist for the Delta Green tabletop roleplaying game franchise. This is the first model that I've trained for SD. I actually trained this one back in December, but never felt like sharing it until now. It does an okay job at capturing the oil painting still of Detweiler's game art. It was trained on SD 1.5, I didn't share it as it had trouble on non-base models. However, I was playing around with it and found that it actually works well with the Vodka model, so I decided that it was high time to release it.
For instructions, I recommend including "oil painting" in your prompt in addition to the TI. It works with base 1.5, and may work with models that were directly trained/finetuned on 1.5. No guarantees that it works on merges or CLIP SKIP 2 models, but I see that there's interest I can go ahead and retrain the concept as a LORA. One thing that this TI can't capture is the low-key lighting in Detweiler's art, so I'd like to try noise offset for it (or train on a model with noise offset enabled).
23/5/28 EDIT: I revisted the dataset and got a LORA version to work. Check it out: The Conspiracy - LORA