Uncensored KREA2 Examples images shows the base model (fp8) on the left / finetune (fp8) on the right (same seed same prompt) The said "FP16" version is the extracted LoRA for standalone usage with any base model version of KREA2, It's in fact BF16 (can't natively post a LoRA on a model page) You can use my "Magic LoRA Loader" from <a target="_blank" rel="ugc" href="https://github.com/PGCRT/CRT-Nodes">https://github.com/PGCRT/CRT-Nodes</a> If you want more precision when tweaking stacks of LoRA Tip - For sharper output, you can use the <a target="_blank" rel="ugc" href="https://huggingface.co/spacepxl/Wan2.1-VAE-upscale2x/blob/main/Wan2.1_VAE_upscale2x_imageonly_real_v1.safetensors"><span style="background-color:rgba(255, 255, 255, 0.03);color:rgb(231, 233, 234);font-family:TwitterChirp, -apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif;font-size:15px">Wan2.1_VAE_upscale2x_imageonly_real_v1</a> with <a target="_blank" rel="ugc" href="https://github.com/spacepxl/ComfyUI-VAE-Utils">https://github.com/spacepxl/ComfyUI-VAE-Utils</a> Loader / decoder Get free credit between 5-500$ in runpod with my link <a target="_blank" rel="ugc" href="https://runpod.io?ref=u7b2habt">https://runpod.io?ref=u7b2habt</a> If you want to support my work. Have fun

Files with SHA256: aab5e174d4a9d56d8bbc83a05e3d52db2e1bc71038a99980107f047f24d95fe2