CelestReal Anima

CelestReal Anima v1.0 — the semi-real line continues on Anima

The spiritual successor to CelestReal. Built on Anima Mayhem and Real Dream — Anima V4, it carries the same anime ↔ realism fusion onto the Anima architecture: the character handling and prompt adherence of a modern anime model, with the skin, light and depth of a semi-real one.

Merge

Anima Mayhem V1

Real Dream — Anima V4 — by @sinatra

Style highlights

• Default look is a soft semi-realism that works across portraits, scenes and NSFW

• Handles dynamic action poses, intimate compositions and cinematic lighting without distortion

• Detailed anatomy and smooth skin rendering

• Character recognition is good, though some characters resist a semi-real read and may come back strange. That is the trade for the look, not a defect.

• Style tuning — one keyword steers the whole look

• Add anime illustration and/or anime → flater, stylised,

• Add realistic and/or photorealistic → lifelike skin and photographic light

• Leave both out → the default semi-real look

• Mixing them fights itself. Pick a side per image.

Required before you run it

This is an Anima family model and it will not load without these two files:

• qwen_3_06b_base.safetensors → ComfyUI/models/text_encoders

• qwen_image_vae.safetensors → ComfyUI/models/vae

Recommended settings

Sampler / scheduler: euler_ancestral + beta (my default; er_sde also behaves)

Steps: 30, then a second pass — this model likes a two-pass workflow

CFG: 5

Resolution: 832×1216 latent → 1248×1824 final

Upscaling

Most of the gallery was made with a plain two-pass workflow and no upscale. Detailers and upscalers only take it further — the look does not depend on them.

If you like what I do and want to support me, you can head over to Ko-fi

Buy WhistlerAi a Coffee