<h1 id="krea-2-gguf-weights"><strong>KREA-2 GGUF Weights</strong></h1><p>Quantized GGUF weights for the Krea-2 Base model, optimized for low-VRAM and consumer hardware local workflows via ComfyUI.<br /></p><h3 id="gguf-model-files:https:huggingface.corealrebelaikrea-2_ggufstreemain"><span style="color:rgb(250, 82, 82)"><strong>GGUF MODEL FILES:</strong><br /></span><a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai/KREA-2_GGUFs/tree/main"><span style="color:rgb(34, 139, 230)">https://huggingface.co/realrebelai/KREA-2_GGUFs/tree/main</span></a></h3><p></p><h2 id="overview"><strong>🚀 Overview</strong></h2><p>Krea-2 is a Diffusion Transformer (DiT). Standard GGUF custom nodes will throw an <code>Unexpected architecture type</code> error natively. <strong>To run these weights, you must use our patched custom node fork.</strong></p><h2 id="downloading-and-installation"><strong>📥 Downloading and Installation</strong></h2><h3 id="1.-custom-node-requirements-(mandatory)"><strong>1. Custom Node Requirements (Mandatory)</strong></h3><p>Standard loaders do not yet recognize <code>krea2</code> architecture tags in GGUF metadata. Clone the following custom node into your <code>ComfyUI/custom_nodes/</code> directory:</p><ul><li><p><strong>Repository:</strong> <a target="_blank" rel="ugc" href="https://github.com/RealRebelAI/ComfyUI-GGUF_KREA-2"><strong><u>RealRebelAI/ComfyUI-GGUF_KREA-2</u></strong></a></p></li></ul><h3 id="2.-required-support-models"><strong>2. Required Support Models</strong></h3><p>Place these base model files in your standard ComfyUI <code>models/</code> directory:</p><ul><li><p><strong>Text Encoder (CLIP):</strong> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen3-VL/blob/8bb77f7ad334bc594cd68c0c48e868abd1e0f719/text_encoders/qwen3vl_4b_fp8_scaled.safetensors"><strong><u>Qwen3-VL-4B-FP8-Scaled.safetensors</u></strong></a> <em>(Load via </em><code>CLIPLoader</code><em>, ensuring Type is set to </em><code>krea2</code><em>)</em></p></li><li><p><strong>VAE:</strong> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main/split_files/vae"><strong><u>qwen_image_vae.safetensors</u></strong></a> <em>(Load via </em><code>VAELoader</code><em>)</em></p></li></ul><h3 id="3.-gguf-weight-installation"><strong>3. GGUF Weight Installation</strong></h3><ol><li><p>Download your preferred quantization file (e.g., <code>Krea-2-Base-Q4_K_S.gguf</code>) from the <strong>Files and versions</strong> tab of this repository.</p></li><li><p>Place the downloaded <code>.gguf</code> file into your <code>ComfyUI/models/unet/</code> (or equivalent GGUF directory) folder.</p></li><li><p>Load the model in ComfyUI using the <code>UnetLoaderGGUF</code> node.</p></li></ol><h2 id="workflow-notes"><strong>⚙️ Workflow Notes</strong></h2><p>Ensure your conditioning setup is mapped appropriately to handle the stacked hidden state requirements of multimodal DiTs to avoid latent sequence mismatch errors.</p>

Files with SHA256: 1e990ce910327fbde128db6631251f6faba604b5c134d668576858be0f56dcff