CitoRooN Mix v1

**An Illustrious merge for anthro, kemonomimi and anime characters that doesn't default to pale skin.** 18+ · pull up a chair.

Citoroon's First model release and more coming! This is my first model share and looking to build up the testing harness and rating system to make merges and model updates much easier and faster. Comments and any advice are welcome!

What we were going for

- One checkpoint for all three worlds. Full anthro furries, kemonomimi, and PoC anime humans, without switching models between them.

- A blend that tries to mix proven models. A merge should beat both of its parents, not land halfway between them. This one does in our tests.

- A default that isn't pale. Most anime and furry models draw a pale woman unless you tell them otherwise. This one leans warm by default, and still goes pale the moment you ask.

- Built by testing, not by feel. Every choice below was rendered and rated 1 to 5 before it made the cut.

The mix

|---|---|---|---|

| Furry and More v6.0 | NexorAi | The body: anatomy, species range, anthro and kemonomimi | 85% of the input and output blocks, 40% of the middle block, 75% of the text encoders |

The VAE is Dixar's, baked in. There are no other LoRAs inside.

For merge nerds, with Dixar as model 1:

IN 0.15 · MID 0.60 · OUT 0.15 · time/label embeddings 0.25 · text encoders 0.25 flat · Skin Color Slider +0.70 · VAE: Dixar

### How we got here

1. The ratio. We rendered seven arms, from pure Furry and More to pure Dixar. Every blend

between 25% and 50% Dixar beat both parents by about a full point: 4.1, against 3.1 and 2.9.

25% was the only blend that held up on fresh seeds.

2. Where Dixar sits. We kept roughly the same share of Dixar and moved it between the layers.

Concentrating it in the middle blocks scored 4.37 against 3.50 for an even spread, across

32 rated images each. It led on anthro, kemonomimi, style and pose alike, and held up on seeds

it had never seen.

3. The skin tone. Skin LoRAs baked in without their trigger words left the default untouched.

A slider moves the default itself. At +0.70 it won our "is this the default we want" rating. At +1.0 it

started darkening fur, so we stopped at 0.70.

How to prompt it

Trigger words: none
- It's a checkpoint, so there's nothing to unlock. Prompt it like any Illustrious model. The skin slider is baked in and needs no words of its own.

Settings

| | |

|---|---|

| Sampler | Euler a |

| Scheduler | Beta |

| Steps | 40 |

| CFG | 5 |

| Size | 832×1216 (what we tested) |

Every rating above was made at these settings. Other SDXL sizes should behave as they do on any Illustrious model.

Also works: DPM++ 2M Karras at CFG ~3.8. On the same seeds it gave the same scenes with a slightly brighter, finer-textured look, and the warm default skin held.

Build the prompt

Tags and plain phrases both work, and our prompts mix them:

masterpiece, best quality, anthro wolf woman, grey fur, adult woman in her 30s, athletic build, leather jacket and jeans, confident smirk, standing, neon-lit city street at night, cowboy shot, rim lighting

- Anthro: write anthro <species> woman (or man), then give a fur colour.

- Kemonomimi: name the ear and tail colour, for example orange fox ears and tail. It anchors

which animal you get.

- Anime humans work too. Skin tone follows the notes below.

Skin tone

- The default leans warm. For pale skin, write pale skin or fair skin; weight it up to

(pale fair skin:1.3) if needed. For deeper tones, write dark brown skin or

deep brown skin.

- Dark tones can bleed into clothes and lips. If a colour matters, name it, for example

teal hoodie or red lipstick.

Negative prompt

Start from:

worst quality, low quality, blurry, jpeg artifacts, deformed, bad anatomy, bad hands, watermark, text, (oversaturated:1.3), (blush:1.2)

The (blush:1.2) follows Dixar's recommendation: its training data runs blush-heavy.

### Adults only

This is an 18+ model. Anchor every character as an adult in the prompt, for example

adult woman in her late 20s or mature adult man in his 30s. We do it in every render.

LoRAs

- Illustrious LoRAs load normally. Start concept LoRAs low. In our tests on an earlier build:

- DredlocksIL by Shrekman17 worked at 0.25 with no side effects.

- At 0.60 it began changing outfits and scenes in prompts that never asked for locs.

- **LoRAs and prompts tuned on another checkpoint may need their weights and settings adjusted

here.**

## Known quirks (v1)

- Warm tones can bleed into clothing and lipstick colours; name the colours you want.

- Light fur can come out a shade greyer than asked. That comes from the skin slider.

## Credits

This merge stands on three creators' work. Thank you.

- Dixar 4: Dix Galore by typhoon2285: https://civarchive.com/models/1456548

Dixar's licence requires credit, so please pass it along if you build on this model.

- Furry and More v6.0 by NexorAi: https://civarchive.com/models/1590959

- Skin Color Slider (Pony / Illustrious) by Topplok2: https://civarchive.com/models/466467

- The base is Illustrious XL by OnomaAI, released under the **Fair AI Public License

1.0-SD**. This merge is shared under the same licence, and its full merge recipe is in "The

mix" above.

Using this model

You're welcome to merge it, build LoRAs on it, and use your generations commercially. Credit this

model, and Dixar: its licence requires credit all the way down the line.

---

CitoRooN AI · original AI · 18+ · pull up a chair.