This is a very straight-forward and simple workflow that I have been using for NSFW generations with Krea 2. (It is a slightly modified version of the stock ComfyUI INT8 Convrot workflow.)<ul><li>This workflow has the "prompt enhancement" portion from the stock Comfy workflow REMOVED. I often use LLMs like Grok for my prompting, it saves time within the workflow</li><li>ReActor node + image input added for face-swaps - bypass these two if unnecessary</li></ul>I would HIGHLY SUGGEST downloading and using the Krea2FilterBypass EVEN WHEN generating SFW content. The bypass greatly increases prompt adherence, along with increasing the natural/emotional feel of life-like photos. It makes everything feel more genuine. Use the 2vector variant.My generations are run at 12 steps using the turbo LoRA listed below at .6 (I've found better results using the Raw model + this LoRA vs the Turbo variant model). Use a combination of Euler - Simple, Denoise 1, and CFG 1 (negative prompt disabled because of this with ZeroConditioningOut node).I set the Krea2FilterBypass anywhere between 1.0-10.0, and I am still experimenting. I've found the best results at 5.0 with a NSFW LoRA (Realism Engine linked below) at .8 strength.Here is a list of models that I use with this workflow:Diffusion Model:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Krea-2/blob/main/diffusion_models/krea2_raw_int8_convrot.safetensors">https://huggingface.co/Comfy-Org/Krea-2/blob/main/diffusion_models/krea2_raw_int8_convrot.safetensors</a></li></ul>Load Clip/Text Encoder:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Krea-2/blob/main/text_encoders/qwen3vl_4b_fp8_scaled.safetensors">https://huggingface.co/Comfy-Org/Krea-2/blob/main/text_encoders/qwen3vl_4b_fp8_scaled.safetensors</a></li></ul>Vae:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_fp32.safetensors">https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_fp32.safetensors</a></li></ul>Here are the 3 LoRAs that I use for my generations (TURBO &amp; BYPASS INCLUDED):<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_turbo_lora_rank_64_bf16.safetensors">https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_turbo_lora_rank_64_bf16.safetensors</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2728234/krea2filterbypass?modelVersionId=3066812">https://civarchive.com/models/2728234/krea2filterbypass?modelVersionId=3066812</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2688234/realism-engine-ideogram-4-krea-2?modelVersionId=3109006">https://civarchive.com/models/2688234/realism-engine-ideogram-4-krea-2?modelVersionId=3109006</a></li></ul>Huge thanks and shoutout to "nsfwVariant" and this awesome guide they posted: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2749367/krea-2-simple-gen-workflow-for-high-quality-realism-lots-of-info-and-tips">https://civarchive.com/models/2749367/krea-2-simple-gen-workflow-for-high-quality-realism-lots-of-info-and-tips</a><ul><li>Be sure to check out their post for some more in depth information on Krea 2 censoring, and the effects that LoRAs/the Bypass have on generating.</li></ul>

Krea 2 - Simple Workflow (NSFW)

First WIP Iteration of my Krea 2 NSFW workflow.

Files with SHA256: 5031946077c49b211880e2ef91df415b59aec21952c4759b16001d53e2ecc8ce