Description
About this version
This should finally fix the wonky looking eyes, and motion distortion is like much much less IMO.
I trained the high up to 117.5K steps! The loss is now average 0.056 which is more than double improvement over V1. I will train another 12 hours and and maybe upload that before I leave for vacation.
I haven't run extensive tests but I think High V1.1 and Low V1.1 both work best together. But do test Low V1 with it and let me know. Also here is some other high epochs if you wanna test them yourself https://huggingface.co/datasets/comfyuiman/gurrenlagannlora/tree/main let me know if any is better. I have not noticed signs of overtraining but I haven't run many tests.
I will be gone for 2 weeks, so I wanted to get this out there so you can try and I'll revisit when I'm back.
Details
Downloads
13
Platform
TensorArt
Platform Status
Available
Created
9/7/2025
Updated
9/7/2025
Deleted
-
Trigger Words:
GurrenLagannStyle.
Files
gurrenlagannstyle_22_High_v2-e298-20250911013557.safetensors
Mirrors
Other Platforms (TensorArt, SeaArt, etc.) (1 mirrors)
Available On (6 platforms)
TensorArt
Wan Gurren Lagann Anime Style - tazmannner379 - T2V V1.1 HIGHSeaArt
Gurren Lagann / Anime Style Wan 2.2 14B Lora - LOW V1.1SeaArt
Gurren Lagann / Anime Style Wan 2.2 14B Lora - HIGH V1.1SeaArt
Gurren Lagann / Anime Style Wan 2.2 14B Lora - HIGH V1SeaArt
Gurren Lagann / Anime Style Wan 2.2 14B Lora - LOW V1SeaArt
Gurren Lagann / Anime Style Wan 2.2 14B Lora - 5B (BETA)