Type | |
Stats | 36 9 |
Reviews | (13) |
Published | Jan 18, 2025 |
Base Model | |
Training | Steps: 8,000 Epochs: 10 |
Usage Tips | Clip Skip: 1 Strength: 0.8 |
Trigger Words | demon_prince_of_Tzeentch Tzeentch, muscular eagle eagle headed wings blue body blue feathers blue skin armor Tzeentchian_armor +15 more |
Training Images | Download |
Hash | AutoV2 63FB02A09C |
This is a Pony lora for the Demon Prince of Tzeentch from the Warhammer fantasy video games. A weight of 0.8 to 1 works well with this. Sampling Steps of 15 to 35 works well with most models, as does a CFG Scale of 3 to 8.
The prompts for this lora are ( demon_prince_of_Tzeentch, Tzeentch, muscular, eagle, eagle headed, wings, blue body, blue feathers, blue skin, armor, Tzeentchian_armor, Tzeentchian_outfit, bejeweled, mark_of_Tzeentch, eye_of_Tzeentch, staff, weapon, tail, chaos_demon, helmet, and hood ). These are for the character.
Mixing up the different costume types should come up with different results. For example: (armor, Tzeentchian_armor, ), (armor, Tzeentchian_outfit, ), and (armor, Tzeentchian_outfit, bejeweled, ). Also using the helmet or hood adds either a full headed helmet or a hood. I only found one example for the hood, so it may not work much.
For the background I used the prompts ( Tzeentchian_chaos_warp, Tzeentch_corruption, Tzeentchian_domain, scenery, Silver_Tower_of_Tzeentch, battle, war, fighting, magic, army, symbol, idol, icon, Asian architecture, Grand Cathay, forest, grassland, grass, desert, pyramids, scenery, Khemri, snow, elven architecture, elven domain, mountain, and hills ) for different background results. The prompts Tzeentchian_chaos_warp and Tzeentch_corruption are the most often used ones to indicate Tzeentchian chaos influence. And Tzeentchian_domain refers to whether the background is in the realm of Tzeentch or not. The rest are for other backgrounds. Simple backgrounds is another common background prompt.
Manually adding prompts to a lora before posting doesn't keep it on there, so I would suggest to just copy and past the prompts you want on the lora by selecting the edit metadata option on Stable Diffusion and then pasting the prompts in the activation text area. Not sure what the deal is with OneTrainer over the metadata thing.
The lora works well on both Pony and Illustration checkpoints.