{"id":16116,"date":"2026-11-17T11:04:27","date_gmt":"2026-11-17T02:04:27","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=16116"},"modified":"2026-11-17T16:12:36","modified_gmt":"2026-11-17T07:12:36","slug":"prunaai-qwen-image-2-1-remove-object","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/prunaai-qwen-image-2-1-remove-object\/","title":{"rendered":"PrunaAI Releases Object Removal LoRA for Qwen-Image-2.1"},"content":{"rendered":"<p><em>Sample outputs are available on the <a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object\">model card<\/a>.<\/em><\/p>\n<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy\">PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy<\/a><\/td>\n<\/tr>\n<tr>\n<td>Family guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-qwen-qwen-image-2-1-en\/\">Qwen-Image-2.1 guide (5 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-alibaba-en\/\">Alibaba (Qwen): models and licenses<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-11-11<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>qwen-research<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">safetensors<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>PrunaAI has released a set of LoRA adapters, <a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object\">PrunaAI\/P-Q-Image-2.1-Remove-Object<\/a>, for the image generation and editing model <a href=\"https:\/\/huggingface.co\/Qwen\/Qwen-Image-2.1\">Qwen\/Qwen-Image-2.1<\/a>. Additionally, <a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy\">PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy<\/a> is available, which includes optimized workflows and dedicated nodes for ComfyUI.<\/p>\n<p>This model is an image-to-image editing tool that removes specified objects from an input image and fills in the background by inferring what lies behind them. By specifying the target for removal via a red-bordered outline or text instructions, the model can remove not only the object itself but also its shadows and reflections on water surfaces, while keeping non-specified areas of the image unchanged.<\/p>\n<h2>Specifications<\/h2>\n<p>The model specifications and recommended settings provided in the documentation are as follows:<\/p>\n<ul>\n<li><strong>Supported Task<\/strong>: Object removal from a single photo (image-to-image)<\/li>\n<li><strong>Model Architecture<\/strong>: LoRA (rank 32)<\/li>\n<li><strong>Input Format<\/strong>: Photo data and removal instruction text (Drawing a red-bordered outline on the photo is recommended; text-only prompts are also supported)<\/li>\n<li><strong>Available Variants<\/strong>: <\/li>\n<li><code>6-step<\/code>: Used with distilled LoRA (<code>PrunaAI\/P-Q-Image-2.1<\/code>) for 6-step generation. Recommended default LoRA weight is 0.7 &#8211; <code>8-step<\/code>: Used with distilled LoRA for 8-step generation. LoRA weight is 0.7 &#8211; <code>base<\/code>: Base model only, 40-step generation. LoRA weight is 1.0<\/li>\n<li><strong>Sampling Settings<\/strong>: No CFG (<code>true_cfg_scale=1.0<\/code>)<\/li>\n<li><strong>Output Resolution<\/strong>: Matches the total pixel count\/area of the input image (the pipeline converts each side to a multiple of 32 for processing, then resizes back to the original image dimensions)<\/li>\n<li><strong>License Terms<\/strong>: <code>qwen-research<\/code> license applies<\/li>\n<\/ul>\n<h2>Performance and Quality<\/h2>\n<p>When attempting object removal with the base <code>Qwen\/Qwen-Image-2.1<\/code> model or its faster distilled versions, issues often arise where background areas other than the target object are altered, or the entire generated image shifts, causing misalignment with the original input photo.<\/p>\n<p><code>P-Q-Image-2.1-Remove-Object<\/code> has been trained to accurately erase the target and its associated effects (such as ground shadows or reflections on wet roads) while preserving the background and maintaining the original spatial relationships.<\/p>\n<p>Quality characteristics and points to note as reported by the model card include:<\/p>\n<ul>\n<li><strong>Linked Removal of Shadows and Reflections<\/strong>: The model automatically infers and removes not only the object specified by the mask but also shadows and reflections extending beyond it.<\/li>\n<li><strong>Texture in Generated Areas<\/strong>: The filled-in area after object removal may appear smoother or blurrier than the surrounding background, lose photo-specific noise, or exhibit regular grid-like patterns. The recommended LoRA weight of 0.7 for the low-step versions is set to mitigate these phenomena.<\/li>\n<li><strong>Masking Precautions<\/strong>: If the mask outline overlaps with people or objects you wish to keep, the outline instruction may take priority and remove adjacent objects as well. In cases where two people overlap, text-only prompts may yield better results.<\/li>\n<\/ul>\n<h2>Strengths and Use Cases<\/h2>\n<p>This model is suitable for restoration tasks such as removing unwanted people, buildings, cars, or equipment from photos and filling the space with a natural background. For example, it excels at removing a person walking on a wet plaza, correctly identifying and removing both the person and the reflections in puddles extending outside the masked area.<\/p>\n<p>It is effective for users who want to use the edited output as raw material, as it prevents the common issue of background distortion or image shifting, maintaining accurate spatial alignment with the original input.<\/p>\n<p>In terms of operation, the recommended method is to draw a red-bordered outline on the photo to specify the target. It also supports execution via prompt instructions alone, such as &#8220;remove the man on the right,&#8221; without creating a precise mask. By combining it with a distilled LoRA, it can be executed in as few as 6 or 8 steps, making it suitable for high-speed image editing workflows.<\/p>\n<p>Note that this model is trained exclusively for object removal tasks and does not support other image editing operations such as adding objects, replacing objects, or changing styles or colors.<\/p>\n<h2>How It Differs from Similar Models<\/h2>\n<p>The specific differences from Qwen-Image-2.1 derivative models previously covered on this site are as follows:<\/p>\n<p><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/09\/30\/qi-2-1-anyangle\/\">\u201cQI_2.1_AnyAngle\u201d Camera Angle Control LoRA<\/a> is designed to change viewpoints and composition while maintaining the original style by referencing images generated from 3D models or Gaussian Splatting. In contrast, this model specializes in removing specified objects and preserving the background while keeping the composition and camera angle completely fixed.<\/p>\n<p>Additionally, <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/09\/23\/viggle-qwen-image-2-1-viggle-turbo\/\">\u201cQwen-Image-2.1-viggle-turbo\u201d 4-step high-speed image generation\/editing model<\/a> is intended for general-purpose image generation and broad, instruction-based editing based on multiple reference images. This model differs significantly in that it focuses solely on &#8220;maintaining the original spatial relationship without changing any area other than the removal target&#8221; and is specifically tuned for handling shadows and reflections.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 7.1B parameters (taken from the base model Qwen\/Qwen-Image-2.1)<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>48GB (RTX 6000 Ada \/ A6000, etc.)<\/td>\n<td>BF16<\/td>\n<td>30.8GB<\/td>\n<td>37.0GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model&#8217;s authors. This release is an adapter (LoRA etc.); the table shows what the base model <a href=\"https:\/\/huggingface.co\/Qwen\/Qwen-Image-2.1\">Qwen\/Qwen-Image-2.1<\/a> needs. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p>The publisher distributes this model as safetensors.<\/p>\n<p><strong>License \u2014 <code>qwen-research<\/code> (Commercial use prohibited):<\/strong> Alibaba&#8217;s Qwen Research License. <strong>Limited to research and evaluation<\/strong>; commercial use requires a separate license (Sections 1.i and 2.b).<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats and the license field. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:lab --><\/p>\n<h2>Our Own Measurements<\/h2>\n<h3>Measurements We Did Not Take<\/h3>\n<p>This model&#8217;s license (<code>qwen-research<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<p><!-- \/lmw:lab --><\/p>\n<p><!-- lmw:files --><\/p>\n<h2>Distributed Files<\/h2>\n<p><em>Weight files published in <a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object\/tree\/main\">PrunaAI\/P-Q-Image-2.1-Remove-Object<\/a>, listed by this site from the Hugging Face API. Sizes are the actual file sizes.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>File<\/th>\n<th>Size<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><code>p_q_image_2.1_remove_object_6_steps_v0.1.safetensors<\/code><\/td>\n<td>168MB<\/td>\n<\/tr>\n<tr>\n<td><code>p_q_image_2.1_remove_object_8_steps_v0.1.safetensors<\/code><\/td>\n<td>168MB<\/td>\n<\/tr>\n<tr>\n<td><code>p_q_image_2.1_remove_object_base_v0.1.safetensors<\/code><\/td>\n<td>168MB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:files --><\/p>\n<h2>How to Get It<\/h2>\n<p>The model weights are distributed in safetensors format. No license agreement (Gated) process on Hugging Face is required for use.<\/p>\n<p>To use it with the Diffusers pipeline, obtain the target LoRA file from the <a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object\">PrunaAI\/P-Q-Image-2.1-Remove-Object<\/a> repository. When running the low-step versions (<code>6-step<\/code> or <code>8-step<\/code>), you must load and apply the distilled LoRA <code>PrunaAI\/P-Q-Image-2.1<\/code> in addition to the base model <code>Qwen\/Qwen-Image-2.1<\/code>. The repository also includes a script, <code>draw_outline.py<\/code>, which draws the recommended red-bordered outline and automatically constructs the prompt.<\/p>\n<p>For ComfyUI, you can set up the environment using <a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy\">PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy<\/a>. This includes LoRA files converted for ComfyUI, ready-to-use JSON workflow files, and a dedicated custom node, <code>precise_removal_outline<\/code>, for handling outlines.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/30\/qi-2-1-anyangle-camera-control-lora\/\">QI_2.1_AnyAngle Camera Angle Control LoRA: 48GB+ VRAM, File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/23\/qwen-image-2-1-viggle-turbo-v0-1-preview\/\">Qwen-Image-2.1-viggle-turbo Image Generation Model: 48GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/qwen-image-2-1-turbo-2\/\">Qwen-Image-2.1-Turbo 8-Step Fast Image Generation Model: 48GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/minimax-h3-image-character-swap-2\/\">MiniMax-H3-Image-Character-Swap Image Editing Model: 4GB+ VRAM<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Explore the same model family<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-qwen-qwen-image-2-1-en\/\">Qwen-Image-2.1 family overview (5 articles, 2 converted builds)<\/a><\/li>\n<li><strong>Formats this model is available in<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">Safetensors format guide and models<\/a><\/li>\n<li><strong>Learn about the publisher<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-alibaba-en\/\">Alibaba (Qwen): models, licenses and articles<\/a><\/li>\n<li><strong>Other models for the same task<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/models-by-task-en\/#task-image\">Other image generation models<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object\">https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy\">https:\/\/huggingface.co\/PrunaAI\/P-Q-Image-2.1-Remove-Object-Comfy<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>PrunaAI has released LoRA adapters for Qwen-Image-2.1 specifically designed for object removal, including ComfyUI workflows and custom nodes.<\/p>\n","protected":false},"author":1,"featured_media":16115,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[203],"tags":[677,2742,111,4945,1835,1547,1839],"class_list":["post-16116","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-image-video-and-audio","tag-comfyui-en","tag-image-to-image-en","tag-lora-en","tag-prunaai-en","tag-qwen-image-2-1-en","tag-verified","tag--en"],"lang":"en","translations":{"en":16116,"ja":16114},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16116","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=16116"}],"version-history":[{"count":3,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16116\/revisions"}],"predecessor-version":[{"id":18263,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16116\/revisions\/18263"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/16115"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=16116"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=16116"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=16116"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}