📋 Description
A merge based on KREA2 Turbo, optimized for fast and high-quality generation. This model combines the strengths of the KREA2 architecture with custom tuning for improved results across various styles.
Three versions have been posted: SAFE (clean art), BASE (nudity allowed), and NSFW (practically anything goes).⚙️ Recommended Generation Settings
Parameter
Value
Steps
8
Sampler
Euler a
Schedule type
Bong Tangent
CFG scale
1
Seed
3448636919
Resolution
1536×2048
Model
fascium_SFWKREA2_fp8
Model hash
98d064dc8f
Module 1
qwen_image_vae
Module 2
qwen3vl_4b_fp8_scaled
Diffusion in Low Bits
Automatic (fp16 LoRA)
RNG
CPU
Style Selector
Enabled (Style: base)
Krea2 Enhancer
✅ Enabled
💡 Tips & Notes
Low CFG (1.0) — This model is designed to work best with a very low CFG scale. Higher values may cause artifacts or oversaturation.
Few steps needed — Only 8 steps are required thanks to the Turbo architecture. Going higher won't improve quality and will just waste time.
Krea2 Enhancer — Make sure to enable the Krea2 Enhancer for best results.
Style Selector — Set to
baseby default. You can experiment with other styles, but disable randomization for consistent output.Resolution — Tested at 1536×2048 (portrait). The model supports various aspect ratios, but keep total pixel count around ~3MP for optimal performance.
fp8 quantized — This is an fp8 version for reduced VRAM usage while maintaining quality.
🖼️ Trigger Words / Prompt Guide
No specific trigger words required. Write natural, descriptive prompts. The model responds well to detailed style and composition descriptions.
⚠️ Known Limitations
SFW-focused merge
Best results with the exact settings above — deviating from recommended CFG/sampler may degrade output quality
Feel free to adjust the model name, add your own example images section, or tweak any details to better match your vision! 🎨
Description
more trained
FAQ
Comments (15)
Any chance to see конврот version?
?
@FASCIUM Имею в виду Новый формат сжатия моделей INT8 ConvRot на примере Krea 2 https://youtu.be/4MMuTYbcE08?si=F5tLij_HeA67vVnF
Вряд ли, я генерирую только в Forge, там эти фичи пока не сделаны. Опять же конечный мердж делаю через чистый скрипт, аналогичная ситуация.
@FASCIUM но тема офигенная, экономит время
An interesting merge. Can you possibly do a semi-raw/semi-turbo version? It generally works like this: the base version is RAW, plus Turbo LoRA at 0.6 weight, plus whatever other loras/embeddings you do on top of that.
Why: because this way it gets MUCH more creative compared to pure Turbo! Like, night and day difference. You still use it at CFG=1, but the number of steps should be 12-14 (up from 8 for pure Turbo). In my opinion, the tradeoff is more than worth it!
I’ve tried RAW, but to be honest, it didn’t really impress me. As I understand it, RAW is good for training but not so much for generation. Then again, working with models is just a hobby for me; once I’m satisfied with the result, I don’t change the models. My business interests lie in a completely different field and take up a lot of time, so I only engage in creative work when I have spare time.
@FASCIUM I was not meaning the RAW raw; I meant the raw checkpoint with the Turbo lora injected at 0.6 weight (at the time of doing the merge). This is currently the best approach for maximum diversity.
Стандартная модель все таки не очень выводит длинные тексты на кириллице. Если еще и с разными регистрами. Судя по вашим примерам - вы как то поправили в своем слиянии? ERNIE, на сколько помню лучше всего спрвляется.
I would like to see more ethnic capability trained into Krea 2. It seems to be blank canvas with a lot of capability. But if you wanted to make a Tamil Indian or other complex genotype my go to is still z-image-turbo.
В целом неплохо. Лица дествительно как с плакатов эпохи застоя, но уж больно не реалистично. Хотелось бы немного меньше пластика
Для реализма базовая модель вполне достаточна
Love this checkpoint, but unsure about v3. While it improves in some areas, the facial expressions are often weirdly theatrical and over the top compared to v2. Still best checkpoint for Krea atm. Thanks!
now this is a cool dragon!
Привет, интересная моделька, что мне понравилось так это разнообразие картинок которое получается на одном промпте, можно не включать спец. ноду, в целом у модели есть свой шарм, она заслуживает внимания. Спасибо что продолжаете радовать своей работой.