Why this model?
I designed Kreation to overcome the limitations of Kreamania. I could no longer expand that model's graphic and painterly capabilities—which had previously fully satisfied me—without compromising its realism.
However, I was unable to achieve specific effects with Kreamania (such as abstract art, Cubism, Symbolism, or stylized drawings). Even when using LoRAs, the shapes remained rigidly tied to realism (for instance, making a square human head was simply impossible).
Rather than attempting to modify Kreamania, I chose to explore a new path with this model, prioritizing an unconstrained painterly and graphic style.
Kreation is intentionally incapable of producing realistic images; its focus is strictly limited to painterly and artistic styles. This way, both approaches can be developed independently.
I'm currently only distributing the optimized inference format (INT8/ConvRot version) to keep performance high and protect the core workflow. The BF16 master weights won't be released publicly for this version. Thanks for understanding!
Description
Steps : 8 - 10
Sampler : er_sde
Scheduler : Simple
Krea2T Enhancer Advanced preferably disabled
The trigger word is not essential.
FAQ
Comments (17)
Excellent work as always! The renders are looking beautiful. Downloading now.
I understand why you don't want to publish BF16, but not everyone has the latest Nvidia graphics cards to use convrot. Perhaps you could publish a more compatible FP8 version.
Int8 compatibility starting from RTX 20xx. Update your comfyui/forge_neo.
@KolAR What's so strange about the fact that people might not have Nvidia cards at all?
Supposedly my set up fully supports int8convrot, but I have "something" misconfigured because all the convrot checkpoints end up horribly noisy for me. I know, that's a "me problem" but still, a non-convrot would be nice and I can understand not sharing bf16 for free. But, eh, I guess there's always non-local gens. :)
@greg241 It is not a question of monetizing models, but rather of highlighting the appropriation of my work (Kreamania) for purely commercial purposes. I already earn a living using AI in my professional life; I am here to have fun, learn, and share. Unfortunately, this sentiment is not shared by everyone in the community
@Adel_AI can you upload FP8 version please?
@Unicom ok, but I'm waiting for my modem to be replaced tomorrow, my upload speed has dropped to almost zero.
Do I need to use Kreafineart Lora with this model, or does this model already include your Kreafineart Lora?
The model remains compatible with all my LoRAs, even those included in the merge. Since the weights used during the merge were low, reusing the LoRAs during inference does not cause any aberrations.
@Adel_AI Okay. Adding your Kreafineart Lora will bring a different style. We can also use it if we want. I understand that. So this applies to your other Lora designs as well. Thank you. I tried the model. It's very artistic.
sorry, I am newbie here, it may sound stupid, but how can I use this model? There is no create button here...
Civitai does not currently offer the option to generate content using the community's fine-tuned KR2 models. I expect this feature to be added soon.
Unfortunately, the only current option is to download the model and generate content locally.
@Adel_AI okay, thaks! Could you tell me how are you generating locally? I figured how to generate in colab using image z turbo, but it has no advansed settings. I really dunno where to start learning :/
@kotanoktan786 I’m not exactly an expert on local generative AI environments. I’ve used ComfyUI from the very beginning; it can be a bit tricky for a beginner since it uses a node-based interface—it takes some getting used to—but if you want to get off to a good start, that’s the platform I’d recommend; it’s the global standard in the field.
You can find some great tutorials on YouTube to get started—I’d recommend the Pixaroma channel for learning the ropes with ComfyUI.
@Adel_AI thanks! I'l try!
@Adel_AI thanks! I'l try!