For image generation, I used the my workflow with some additional custom nodes, for a upscaler feel free to use any upscaler you prefer.I went with the sampler combination er_sde/simple. While some other combinations can produce even better results (exp_heun_2_x0 + sgm_uniform , euler + beta ), experiment and maybe YouI’ll get better results with other samplers ( some of the pictures I also worked with other samplers ). The recommended model strength ranges from 0.90 to 1.0, but feel free to experiment with other values in this case I used most 1.0. The number of steps is the standard for Krea 2 but you can go even higher, though you can of course adjust that as well. The prompt must be detailed with all important details , a prompt from a few words or two sentences will not bring the expected results (For the first three examples I posting in the showcase, I also including and the prompt template I used ), you can extract various styles from the model ,black and white,noir and of course, the Frank Miller Sin City effect ! In this case I used only this model but in experimenting I found that he show great results with this three models . In case Krea censors the image, I recommend:

1. Krea2 TextFusion Refusal-Reduction LoRA https://civarchive.com/models/2775340/krea2-textfusion-refusal-reduction-lora?modelVersionId=3125118;

2. For a stronger effect, the KREAL Cinematic model https://civarchive.com/models/2825469/kreal?modelVersionId=3254511;

3. For more detail, the [KREA 2] Detail Slider https://civarchive.com/models/2729908/krea-2-detail-slider?modelVersionId=3068874