Not loaded
Download
1 variant available
fp32 SafeTensor
ThorGOW_Krea2.safetensors
Full precision, largest file • 447.57 MB
Verified: 4 days ago
Thor from God of War Ragnarok
Been working on a good Thor Lora for 2 years now but was never satisfied with the results - he is really hard to train given the limited references as well as limitations of current image models.
Getting his clothes, tattoos and facial features exactly right is super hard. So you will see slight difference in patterns on his clothes/tattoos but I simply don't have the patience anymore to get it 1:1 right. Should still be the most accurate Thor Lora out there.
For wider shots/full body shots you will have to go really high res (2048) or use a eye detailer to fix his eyes/face. Again its because he has so many distinct facial features that current image models simply cannot cram all of it in a small pixel space.
Lora isn't perfect by any means but I will wait for a better model that supports proper higher res training (Krea2 "caps out" at 1024) to re-train it.
Show more

5260 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
1.2K0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
Main activation tag
Thor the Norse warrior god.Main outfit (top to bottom)
wearing brown leather shoulder pauldrons, teal cloth forearm wraps, brown leather bracers, a ring necklace, brown leather faulds, multiple leather belts, loose dark teal trousers, brown leather boots with crisscrossed lacing, gold rings on several fingers, with a large ornate war hammer at his side
Recommended samplers:
- [Krea 2 Turbo] Euler Scheduler: Simple CFG: 1.0 Steps: 8 Lora strength: 1.0 - good for stylized
- [Krea 2 Turbo] Euler Scheduler: Beta CFG: 1.0 Steps: 8 Lora strength: 1.0 - maximum realism
Prompt tips (green - positive, red - negative):
- Full-body shot / medium shot / close-up portrait / medium-full shot / cowboy shot - example of shot types used during training
- Photorealistic render / 3D render - default Lora style. For other styles use descriptive prompts (examples in gallery)
Last image in the gallery can be imported into ComfyUI with a basic workflow:
No custom nodes required.


