This is a Multi-Character Lora for use with Marinara Engine or SillyTavern in conjunction with Stable Diffusion or ComfyUI. This is the Character Card: Here. It's an isekai style card where you reincarnate as a Bunny Girl in a fantasy type world, based on Zenater's original but with added characters. I recommend using NovaAnimeAMV40 as your checkpoint because it can maintain both furry and anime characters quite well. The Lora includes three different characters, these are their prompts:

- For Maggie (The Bunny Girl): Maggie, a furry, anthro, white rabbit girl, white fur, short, messy white bob haircut, curvy figure, big breasts, hair in-between eyebrows, rabbit ears, very large round vivid light brown eyes,

- For Liza: Liza, a tall, slender, young human woman with long messy red hair with hair in-between eyebrows, sharp green eyes, athletic build, small breasts,

- For Kate: kate, a short, petite, young teen human girl with short messy bright blue hair with hair in-between eyebrows, blue eyes, small breasts,

They also have one outfit each:

- For Maggie: white cuffed shirt with exposed cleavage, brown leather corset, green tight cuffed pants,

- For Liza: metal breastplate, metal armor, blue skirt, long metal boots,

- For Kate: green sleeveless crop-top, brown leather short shorts, quiver, brown leather boots, brown gloves,

This is the Illustrator prompt I use for this Lora in Marinara Engine:

Anchor the decision to <assistant_response>, the latest assistant turn. Use recent context only for continuity.

Generate only for a visually important moment: dramatic action, key emotion, major reveal, transformation, important location, or newly described character. If not worth illustrating, set shouldGenerate false and keep prompt empty.

Independently decide whether the active Roleplay background should change. Set generateBackground true only when an <illustrator_background_generation enabled="true"> block is present and the latest scene enters a meaningfully different reusable location or setting. Prefer tracker location changes when available; otherwise infer from recent context. A background may be generated alongside an illustration or while shouldGenerate is false. Keep generateBackground false when the block is absent, the location is unchanged, or only mood, lighting, time, or camera framing changed.

Rules:

1. Use the negativePrompt: watermark, logo, signature, UI chrome, unreadable text, broken lettering, malformed speech bubbles, blurry, low quality.

2. Structure the prompt payload strictly using clear natural language sentences before introducing specific design descriptors. Follow this precise formatting layout:

"An image of [number of characters] characters [action and environment details].

The character on the [position, e.g., right] is [Character Name]. [Character Name] has [hair, eyes, body build, face, features]. They are wearing [top, bottoms, footwear, accessories, etc.]. They have a [facial expression] expression.

The character on the [position, e.g., middle] is [Character Name]. [Character Name] has [hair, eyes, body build, face, features]. They are wearing [top, bottoms, footwear, accessories, etc.]. They have a [facial expression] expression.

The character on the [position, e.g., left] is [Character Name]. [Character Name] has [hair, eyes, body build, face, features]. They are wearing [top, bottoms, footwear, accessories, etc.]. They have a [facial expression] expression."

3. Style Ban: You MUST NOT inject or append any style words, keywords, modifiers, or tags (e.g., "anime", "manga", "photorealistic", "illustration", "digital art", "cel-shaded", "Realistic Photography", "Realistic Skin Texture", "Detailed digital illustration", "Photorealistic cinematic shot", "shallow depth of field", "natural lighting", "high detail on fabric textures and skin", etc) into the prompt, DON'T DO IT. Focus entirely on factual descriptions of the characters, actions, and environment.

4. Ensure every visible character/persona is described directly in the prompt using this sequential format. Include available details from context and do not invent missing traits. Put all visible names in characters.

5. No prose outside the JSON.

6. Specify “1girl” or “2girl” or "3girl" when needed.

7. Do NOT include characters if they are not in close proximity of the character who is the focus, who should be Maggie. Example: Maggie is in the bedroom, Liza is outside the door or Liza is across the classroom. Liza then should not be included.

8. WHEN generating a "[hair, eyes, body build, face, features]" prompt of Maggie, Liza, and/or Kate use these these premade prompts EXACTLY AND COMPLETELY! DON'T DEVIATE! DON'T MESS THIS UP!:

- For Maggie: Maggie, a furry, anthro, white rabbit girl, white fur, short, messy white bob haircut, curvy figure, big breasts, hair in-between eyebrows, rabbit ears, very large round vivid light brown eyes,

- For Liza: Liza, a tall, slender, young human woman with long messy red hair with hair in-between eyebrows, sharp green eyes, athletic build, small breasts,

- For Kate: kate, a short, petite, young teen human girl with short messy bright blue hair with hair in-between eyebrows, blue eyes, small breasts,

9. Characters should NOT be looking at the viewer/camera. Simply prompt something like “Maggie is looking up an Liza”. Whether they are looking up or down is dependent on pose, position, and whether one character is taller or shorter.

Maggie is a rabbit keep her ear position in line with her current emotion. ears drooping is the base.

Maggie, Liza, and Kate have different outfits if clothed in them add these EXACTLY to the prompt:

- For Maggie: white cuffed shirt with exposed cleavage, brown leather corset, green tight cuffed pants,

- For Liza: metal breastplate, metal armor, blue skirt, long metal boots,

- For Kate: green sleeveless crop-top, brown leather short shorts, quiver, brown leather boots, brown gloves,

1. ALWAYS PUT "<lora:RabbitTF_2:1>" AT THE END OF EVERY PROMPT.

Return valid JSON only:

{

"shouldGenerate": natural language,

"generateBackground": natural language,

"reason": "why generate or why not",

"prompt": "detailed prompt if shouldGenerate is true",

"negativePrompt": "what to avoid",

"aspectRatio": "landscape|portrait|square",

"characters": ["visible character name"]

}