Branwen lora, using a dataset i've been refining since Illustrious/Noob. Wanted to share it because every other lora for her on here is complete garbage and can't even get her basic outfit right, let alone her tattoo. This one does everything including the tattoo pretty accurately, aside from some upside down/underleg angles.

Trained on mostly anime screenshots and some pictures from her official artbook by Oda Non (supports that artbook style via '@oda non', but probably needs a dedicated lora for that style to really capture it properly).

Highly recommend using the anima turbo v0.1 lora alongside this - it makes the images look better/more stable and is way faster to run. The examples were all using v0.1 of the turbo lora.

Style tags to get the 'anime' look from the examples:

queen's blade, anime screenshot

Outfit tags are optional, but to force (or block via negative prompt) pieces of her outfit:

collar, corset, thong, sandals, red armbands, ass tattoo, leg tattoo

Adding "breasts out", "nude", "topless", or "bottomless" can also be used to remove parts of the outfit.

'Dungeon' setting from her OVA:

dark, dungeon background, sidelighting

For Dogura:

Dogura, goblin

(or "small goblin" if it's making him too large in the shot).

Trained on both natural language and tag variants of captions.