<h1 id="refcontrol-flux.2-klein-4b-reference-depth-lora">RefControl FLUX.2 Klein 4B - Reference Depth LoRA</h1><h2 id="short-description">📝 Short description</h2>A LoRA for FLUX.2 Klein 4B Base that fuses a reference image (left) with a depth map (right). It preserves identity and style from the reference while following the pose and structure from the depth map. Model: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2655452/workflow-refcontrol-flux2-klein-9b-reference-depth-lora?modelVersionId=2981746">https://civarchive.com/models/2655452/workflow-refcontrol-flux2-klein-9b-reference-depth-lora?modelVersionId=2981746</a>Trigger word: <code>refcontrol</code><hr /><h2 id="examples">📊 Examples <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c9c4e7ab-2dca-4f63-b6fb-e29391a081cf/original=true/c9c4e7ab-2dca-4f63-b6fb-e29391a081cf.jpeg" /></h2><h2 id="extended-description">📖 Extended description</h2>This LoRA was primarily trained on humans, but it also works with objects and stylized characters. Its main purpose is to preserve identity - facial features, clothing, style, or object characteristics - from the reference image, while adapting the subject to the pose and composition defined by the depth map.Part of the RefControl family: reference + control fusion for consistent, controllable generation on FLUX.2 Klein 4B Base.<hr /><h2 id="how-to-use">⚙️ How to use</h2><ol><li>Use first image as depth reference, second image input.</li><li>Add the trigger word <code>refcontrol</code> in your prompt.</li><li>Adjust LoRA weight (recommended 0.8-1.0) depending on how strongly you want to preserve identity.</li></ol><h2 id="comfyui-requirenments">ComfyUI requirenments</h2>For extracting depth you need to install nodes with depth estimation in workflow used:<a target="_blank" rel="ugc" href="https://github.com/Fannovel16/comfyui_controlnet_aux">https://github.com/Fannovel16/comfyui_controlnet_aux</a><h3 id="base-model">Base model</h3>Use with <code>black-forest-labs/FLUX.2-klein-base-4B</code>. The undistilled Base variant is intended for LoRA training and custom pipelines (~50 inference steps, <code>guidance_scale</code> ~4.0).<h3 id="example-prompt">✅ Example prompt</h3><pre><code>refcontrol
</code></pre><hr /><h2 id="what-it-does">🎯 What it does</h2><ul><li>Preserves character or object identity across generations.</li><li>Embeds the subject into a new pose or scene defined by the depth map.</li><li>Works best when the depth map has similar proportions and scale to the reference.</li></ul><hr /><h2 id="tips">⚡ Tips</h2><ul><li>Best results when the depth map is not drastically different in object scale from the reference.</li><li>Combine with text prompts to refine background, lighting, or mood.</li><li>Can be combined with other RefControl LoRAs (pose, lineart, canny) on compatible pipelines for multi-constraint generation.</li></ul><hr /><h2 id="use-cases">📌 Use cases</h2><ul><li>Human portraits in different poses.</li><li>Consistent character design across multiple scenes.</li><li>Object transformations (cars, furniture, props) with depth-guided placement.</li><li>Storyboarding, comics, or animation frame generation.</li></ul><hr /><h2 id="files">📦 Files</h2><ul><li>Weights: <code>flux2_klein_4b_refcontrol_depth.safetensors</code></li><li>ComfyUI workflow: <code>image_flux2_klein_image_edit_4b_refcontrol_depth.json</code></li><li>Repo: <a target="_blank" rel="ugc" href="https://huggingface.co/thedeoxen/refcontrol-FLUX.2-klein-4B-reference-depth-lora">thedeoxen/refcontrol-FLUX.2-klein-4B-reference-depth-lora</a></li></ul>

Files with SHA256: 389d04f2b2cc8e45c2420fe599e99443fd0de39cfb856f19494f3077c6d83903