Please share a render using this style down below, thanks for downloading!

Rias_HSDXD_Style LoRA

A style LoRA for Anima base-v1.0 focused on generating High School DxD inspired mobage card art, particularly featuring Rias and does Akeno well too due to Danbooru.

Quick Settings (ComfyUI)

• Checkpoint | anima-base-v1.0

• Resolution | 896x1168 (or similar ~1MP)

• VAE | qwen_image_vae

• Text Encoder | qwen_3_600m

• Sampler | er_sde

• Scheduler | sgm_uniform

• Steps | 40-50

• CFG Scale | 4

• LoRA Strength | 1.0 (Default)

• Seed | Randomize

Prompting Guide

1. Structure

Follow the recommended Anima tagging order for best results:

[Quality/Meta/Safety] → [Character Count] → [Character] → [Series] → [Artist] → [General Tags] → [LoRA Triggers] → [Natural Language Context]

2. Triggers

* Primary: rias_hsdxd_style, dainty_anime_hands

* Usage: Start your natural language description with "a rias_hsdxd_style illustration".

* Hands: The model was trained heavily on "peace signs" and "boob grabs." If generating hands, try replacing the word "hand" with this trigger. Results may vary; test with and without.

3. Example Prompt

masterpiece, best quality, score_7, sensitive, year 2026, 1girl, rias gremory, high school dxd, red hair, school uniform, smiling, double v, both dainty_anime_hands showing a double peace sign, A rias_hsdxd_style illustration of Rias holding a book and sitting a desk in the club room with bookshelves in the background.

4. Negative Prompt

Use standard Danbooru-style negative tags.

worst quality, low quality, score_1, score_2, score_3, bad hands, bad anatomy, extra digits, anatomical nonsense, artistic error, extra fingers, text, signature, bad feet, extra toes, censor, blurry, watermark, artist name

Note: I don't recommend using negative prompts to remove core concepts for any of the LoRA's in your stack.

Additional Tips

* Clothing/Poses LoRAs: Compatible with various clothing LoRAs at 0.8 strength, though thanks to Anima, Danbooru tags often suffice.

* Token Count: Keep prompts under 200 tokens for optimal performance. You can estimate this is about 800 characters.

* Character Variety: Works best with Rias and Akeno currently. I plan to expand the dataset to include Asia, Ravel, Rosweisee, Koneko, Sarafall, Irina, Xenovia, and any others that are in the mobage card style in my next Anima LoRA.

* Compatibility: Designed specifically for the Anima architecture; not intended for SDXL or Pony derivatives.