<h1 id="krea-r-turbo">Krea-R-Turbo</h1>Link to Model: <a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai/Krea-R-Turbo/tree/main">https://huggingface.co/realrebelai/Krea-R-Turbo/tree/main</a>Test the model on HF Spaces for free: <a target="_blank" rel="ugc" href="https://huggingface.co/spaces/realrebelai/Krea-R-Turbo">https://huggingface.co/spaces/realrebelai/Krea-R-Turbo</a>Full Weight and GGUF quants of custom merge of Krea-2-Turbo and 2 of my style LoRAs at specific strength values. Displays a heavy focus on photorealistic portraits to achieve high grade aesthetics while also retaining Krea-2s sharp detail!This is a derivative of <code>krea/Krea-2-Turbo</code>: LoRA-merged, then quantized. It is not the original checkpoint. QuantsINT8 Covrot now available!Pick by how much VRAM you can spare for the diffusion model (leave room for the text encoder + VAE). Sizes are approximate — see the Files tab for exact numbers.FileType~SizeNotes<code>Krea-R-Turbo-Q8_0.gguf </code>Q8_0~14 GB Near-lossless. For 16 GB+ cards or offload.<code>Krea-R-Turbo-Q6_K.gguf</code>Q6_K~11 GBExcellent quality, hard to tell from Q8.<code>Krea-R-Turbo-Q5_K_M.gguf</code>Q5_K_M~9.4 GBGreat quality/size balance.<code>Krea-R-Turbo-Q4_K_M.gguf</code>Q4_K_M~8.0 GB Recommended all-rounder. 12 GB cards, or 8 GB with offload.<code>Krea-R-Turbo-Q3_K_M.gguf</code>Q3_K_M~6.4 GBBest pick for 8 GB VRAM. Small quality trade.<code>Krea-R-Turbo-Q2_K.gguf</code>Q2_K~5.5 GBTightest fit. Visible degradation but usable.On an 8 GB card (e.g. RTX 3070), Q3_K_M or Q2_K fit most comfortably alongside the VAE and text encoder. Larger quants still run via ComfyUI's <code>--lowvram</code> / block-swap offload — just slower.<hr /><h2 id="usage-in-comfyui">Usage in ComfyUI</h2>Place the files:FileFolder<code>Krea-R-Turbo-QX.gguf</code> (this repo)<code>ComfyUI/models/unet/</code> (or <code>diffusion_models/</code>)<code>qwen3vl_4b</code> text encoder<code>ComfyUI/models/text_encoders/qwen_image_vae.safetensorsComfyUI/models/vae/</code>The text encoder (Qwen3-VL 4B) and VAE (Qwen-image VAE) are the standard Krea 2 companions and can be pulled from the ComfyUI repack: <code>Comfy-Org/Krea-2</code>.Wire it up: load the GGUF with the Unet Loader (GGUF) node, feed it the Qwen3-VL text encoder and Qwen-image VAE, and sample as normal Krea 2 Turbo.<h3 id="recommended-sampling">Recommended sampling</h3>Krea 2 Turbo is a distilled, few-step model:<ul><li>Steps: ~8</li><li>Guidance: CFG-free — set CFG to <code>1.0</code> (no negative prompt), or guidance <code>0</code></li><li>Sampler / scheduler: <code>er_sde</code> + <code>simple</code> works well (or <code>euler</code> + <code>simple</code>)</li><li>Resolution: up to 2048², 1024–1280 sweet spot on 8 GB</li></ul>No positive/negative CFG split is needed; this checkpoint was distilled without it.<hr /><h2 id="requirements">Requirements</h2><ul><li>ComfyUI (recent build with Krea 2 support)</li><li><a target="_blank" rel="ugc" href="https://github.com/RealRebelAI/ComfyUI-GGUF_KREA-2">RealRebelAI/ComfyUI-GGUF_KREA-2</a> custom node</li><li>Qwen3-VL 4B text encoder + Qwen-image VAE (from <code>Comfy-Org/Krea-2</code>)</li></ul>This repo contains the diffusion model (transformer) only — it will not generate anything on its own without the VAE and text encoder above.<hr /><h2 id="license-and-attribution">License &amp; attribution</h2>These weights are a derivative of Krea 2 Turbo © Krea.ai, Inc., and remain governed by the Krea 2 Community License Agreement and the Krea Acceptable Use Policy — exactly as the original model. See the license: <a target="_blank" rel="ugc" href="https://huggingface.co/krea/Krea-2-Turbo/blob/main/LICENSE.pdf">LICENSE.pdf</a>. In the event of any conflict, the Krea Acceptable Use Policy and Krea 2 Community License control.Deployer obligation: if you serve or redistribute this model, the license requires you to implement content-filtering or equivalent review to prevent generation or distribution of unlawful or policy-violating content. Report harmful or illegal outputs to <code>safety@krea.ai</code>.Upstream:<ul><li>Base model: <code>krea/Krea-2-Turbo</code> (distilled from <code>krea/Krea-2-Raw</code>)</li><li>ComfyUI repack (VAE / text encoder): <code>Comfy-Org/Krea-2</code></li><li>GGUF tooling based on city96's <a target="_blank" rel="ugc" href="https://github.com/city96/ComfyUI-GGUF">ComfyUI-GGUF</a></li></ul>The merge and quantization are RealRebelAI's work; the underlying model rights belong to Krea.ai, Inc.<hr /><h2 id="credits">Credits</h2>Merged, quantized, and released by RealRebelAI.<ul><li>GitHub: <a target="_blank" rel="ugc" href="https://github.com/RealRebelAI">github.com/RealRebelAI</a></li><li>Hugging Face: <a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai">huggingface.co/realrebelai</a></li></ul>If these are useful, a like on the repo helps others find them. Issues with loading go on the <a target="_blank" rel="ugc" href="https://github.com/RealRebelAI/ComfyUI-GGUF_KREA-2">ComfyUI-GGUF_KREA-2</a> tracker.

Files with SHA256: f6077c5a4567a96a653feaa2dde55afdeb162626d3aa13951d61f79f160021b7