We rebuilt the trainer. Watch your run as it trains, generate with an epoch before it's done, and compare epochs side by side. It's in beta, and you switch it on yourself in your preferences.

Training Studio: Watch It Train, Generate As It Goes

A training and content contest with over 2 Million Buzz in prizes and 50% off H3 generation for the duration of the contest.

The Civitai H3 Contest: 2 Million Buzz Up For Grabs

<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/49ef4f64-5bbf-4a88-9105-6c8acb1e2a37/original=true/49ef4f64-5bbf-4a88-9105-6c8acb1e2a37.jpeg" /> Files:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/vantagewithai/Krea-2-Raw-GGUF/blob/main/krea2_raw-Q6_K.gguf">Diffusion model</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/ethanfel/Qwen3-VL-4B-Heretic-ComfyUI-ConvRot-INT8/blob/main/qwen3-vl-4b-heretic_int8_dynamic_convrot.safetensors">Text encoder</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_fp32.safetensors">Vae</a></li><li><a target="_blank" rel="ugc" href="https://civitai.red/models/2726698/krea-2-raw-turbo-extracted-lora?modelVersionId=3064924">Turbo LoRA</a></li><li>If you want ControlNet, you need to git clone the <a target="_blank" rel="ugc" href="https://github.com/facok/comfyui-krea2-controlnet">repo </a>into custom_nodes and get <a target="_blank" rel="ugc" href="https://huggingface.co/Patil/Krea-2-depth-controlnet/tree/main">this LoRA model</a>.</li></ul>Ollama model:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/mradermacher/gemma-4-12B-it-uncensored-heretic-GGUF/blob/main/gemma-4-12B-it-uncensored-heretic.Q8_0.gguf">https://huggingface.co/mradermacher/gemma-4-12B-it-uncensored-heretic-GGUF/blob/main/gemma-4-12B-it-uncensored-heretic.Q8_0.gguf</a></li></ul>Enjoy!

Bugfixes, small adjustments to ollama system prompt for better NSFW. Reduced the number of needed LoRAs to minimum.

Optimized the nodes including ollama prompt gen.

Takes a bit more time for even more realism and correct posing.

Tested a whole bunch of LoRAs and combos.. small changes.

New samplers/schedulers for more realism, swapped to Krea 2 raw with turbo LoRA at 0.6 weight, found a nice ratio for other LoRAs, links in image metadata.

Changed workflow to support ollama prompt enhancement.

Added 2nd pass convenience for more realism and speed of iteration, and prompt enhancer.

Blink's Krea 2 simple photo-realistic workflow | ti2i 16GB VRAM CivitAI metadata AIO

blinkdothleh.png

wide hips

downblouse

pg-13

suggestive

oral invitation

pg13

sexy

huge breasts

thick thighs

physical violence

weapon violence

revealing clothes

convenient censoring

corpses

male nudity

female swimwear or underwear

female nudity

breast out

male underwear

hair over breasts

female swimwear

gigantic breasts

graphic violence or gore

emaciated bodies

nude

lingerie

sexual situations

disturbing

male swimwear or underwear

partial nudity

undressed

breasts out

exposed female nipple

no panties

covered nipples

huge butt

strapless leotard

sitting on face

one breast out

female underwear

nsfw

graphic female nudity

hentai

porn

genitals

graphic male nudity

adult toys

illustrated explicit nudity

nudity

futanari

sexual intent

peeing

sexual activity

incest

vore

oral

anal

blowjob

dildo riding

hanging

hate symbols

nazi party

scat

extremist

child on child

white supremacy

diapers

self injury

hate speech

urine

latex clothing

swimwear

bukkake

fellatio

cumshot

implied fellatio

eat_cum

cumdrip

cum in pussy

cum on face

after fellatio

cum on hair

cum on body

cum on tongue

cum on hands

cum in mouth

triple fellatio

autofellatio

fucked silly

cum on pussy

Tool Creators