Sign In

Elena (Star Ocean 6)

41
266
49
10
Verified:
SafeTensor
Type
LoRA
Stats
69
29
8
Reviews
Published
Jan 7, 2025
Base Model
Pony
Usage Tips
Clip Skip: 2
Strength: 0.9
Trigger Words
elena_so6, clothing cutout, cleavage cutout, fingerless gloves, belt, black pants, armor, highleg, yellow shirt, zipper, shoulder armor, boots
Hash
AutoV2
36A1A9B6A9
Second Birthday Badge
mesnon's Avatar
mesnon

Elena (or Elena E-014297) is a humanoid android from Star Ocean 6. Apparently her looks are designed to "avoid inciting feelings of discomfort". Well, works for me. :D

I decided to make another Star Ocean lora, since the series does have some very nice looking characters, of which people might want to generate more art, as there really isn't much of it at the moment.

Only tested this Lora on 2d checkpoints, but the dataset has quite a bit of material from in-game cut scenes, so 3d and realistic models should handle the Lora decently well. I preferred using this Lora with 0.5 or 0.6 strength, the higher strengths really seem to mess her hands up for me, but otherwise work decently too.

Now go and unleash the comforting blonde androids into the world!

UPDATE:
PonyXL lora available. NOTE: Due to poor training & low quality dataset, the lora will most likely produce low quality hands. "detailed 3d" prompt gives 3d-styled, decent hands most of the time, but "source_anime" will most likely produce hands that you'll need to inpaint.

I might return to this lora and make another attempt if I somehow can acquire better dataset than what I currently possess.

You can try lowering lora weight to 0.8 and increase the character prompt weight ("elena_so6") to somewhat combat the poor hand quality.

UPDATE 2:

PonyXL version 2 available. Lora should now deal better with generating hands. Uploaded comparison pictures of v1 versus v2. Judge for yourself whether you think V2 is better than V1. Note that the seed and prompt used in the comparison pictures are identical, so the V2 generated picture is what you would get by using the same prompt to generate the pictures as V1. Finally, the first set of uploaded pictures in V2 are my refined prompts from what I used to generate V1 pictures.