First publish. Rook my fursona. 🐶 Be nice :3

What do?

• Go crazy, have fun!

• Make him eat spaghetti. 🍝

• Draw him in a maid dress.

• Draw him doing hot stuff with your character! :3

• etc.

How do?

All the images have simple ComfyUI workflows embedded.

Example prompt (natural language):

Rook_Kaefer. Adult male Felkin resting on a moss-covered fallen tree in an ancient forest, red fur with black ladybug spots, musclegut physique, tail draped over the log, dappled sunlight through leaves, calm smile, photorealistic fantasy wildlife portrait, highly detailed fur and forest textures

Example tag prompt (Noob / e621-ish tags):

masterpiece, best quality, newest, Rook_Kaefer, felkin, male, solo, adult, nude, erect penis, stepping out of shower, shower stall, bathroom, steam, wet fur, come-hither gesture, flirtatious smile, red fur, black spots, red eyes, facial markings, musclegut, dripping tail, tile floor, warm light, sensual, photorealistic, detailed fur

Models

NoobAI - [Ok!]

• v1.5

  • Base Model: NoobAI-XL EPS-pred

  • Superior compatibility with all Noob-based models!

  • Use this to preserve base model style, e.g. Willy‘s Realism v1 Core

  • Ok with t2i

  • Character Inconsistencies: black hands, white fur, white chin, etc.

• v1.0

  • Base Model: Willy‘s Realism v1 Core

  • Wait, SDXL can do THAT..?!

  • t2i and i2i

  • Lora Strength: 1.0 💪

  • DPM++ 2M SDE & beta @ 50 steps, 4 CFG
    

Anima - [Best so far!]

• v1.0

  • Base model: Anima

  • Awesome model, easy to train, easy inference!

  • Expect Finger artifacts!

  • t2i and i2i

  • Lora Strength: 1.0 💪

  • Default samplers/scheduler @ 50 steps

Flux2 Klein 4B/9B - [For Edits!]

• v1.0

  • Base model: Flux2 Klein 4B/9B Base

  • It was hard to teach this model what a cock is !

  • Lora Strength: 0.9-1.0

  • t2i worked best with base model + res_multistep @ 40 steps, 5 CFG.

  • edits work with non-base (distilled 4 step) @ 4 steps, 1 CFG

Chroma - [not so good]

• v0.6

  • Base Model: Chroma1-HD

  • Help! 😣 I suck at chroma.

  • Hard to train, hard for inference. But for some cases, it's awesome.

  • LoRA sucks at t2i, BUT it's still useful for i2i!

  • Lora Strength: ??

  • Sampler/scheduler: ?? (Never found a favorite one)

  • If someone has a workflow that works reliably, let me know!