{"id":129,"date":"2026-07-17T08:00:00","date_gmt":"2026-07-17T08:00:00","guid":{"rendered":"https:\/\/backendside.com\/blog\/?p=129"},"modified":"2026-07-17T09:04:40","modified_gmt":"2026-07-17T09:04:40","slug":"imagelabs-2-nine-on-device-ai-models","status":"publish","type":"post","link":"https:\/\/backendside.com\/blog\/2026\/07\/17\/imagelabs-2-nine-on-device-ai-models\/","title":{"rendered":"ImageLabs 2.0: Nine On-Device AI Models for Background Removal, Upscaling, Colorization and Repair"},"content":{"rendered":"<p class=\"lead\">Version 1.0 of ImageLabs was a tidy, task-based Windows editor with one trick up its sleeve: an offline AI background remover. Version 2.0 is a different animal. It ships a <strong>managed library of nine on-device AI models<\/strong> covering background removal, click-to-select object editing, 4&times; super-resolution, photo colorization and scratch repair &mdash; and, true to the original promise, every last one of them runs on your own machine. No uploads, no accounts, no cloud round-trips. Here is the full tour.<\/p>\n<h2>Five models for background removal &mdash; and four things to do with the subject<\/h2>\n<p>Background removal is no longer a single button. ImageLabs 2.0 gives you a spectrum of open-source segmentation models so you can trade speed for quality as the job demands:<\/p>\n<ul>\n<li><strong>U&#178;-Netp<\/strong> &mdash; the fast, bundled model (4.7&nbsp;MB). It works the instant you install the app, no download required.<\/li>\n<li><strong>Silueta<\/strong> &mdash; a balanced middle ground (43&nbsp;MB) when the tiny model isn&#8217;t quite clean enough.<\/li>\n<li><strong>U&#178;-Net<\/strong> &mdash; the best all-round quality (176&nbsp;MB).<\/li>\n<li><strong>U&#178;-Net Human<\/strong> &mdash; tuned specifically for people and portraits (176&nbsp;MB).<\/li>\n<li><strong>ISNet<\/strong> &mdash; the sharpest, crispest cut-out edges (176&nbsp;MB), ideal for hair and fine detail.<\/li>\n<\/ul>\n<p>Once the subject is detected, the new <strong>Background<\/strong> task offers four modes instead of a plain cut-out:<\/p>\n<table style=\"width:100%;border-collapse:collapse;margin:.5rem 0 1rem;font-size:.9rem;\">\n<thead>\n<tr style=\"background:#f4f6fb;\">\n<th style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:left;\">Mode<\/th>\n<th style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:left;\">What it does<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\"><strong>Remove<\/strong><\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Cut the subject out to transparency. Add a sticker outline, a drop shadow, or trim tightly to the subject.<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\"><strong>Blur<\/strong><\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">A portrait\/bokeh look &mdash; subject sharp, background softly blurred.<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\"><strong>Color pop<\/strong><\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Keep the subject in colour while the background turns greyscale.<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\"><strong>Replace<\/strong><\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Drop the subject straight onto another photo.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>Because the AI mask is cached per image, the model runs <strong>once<\/strong> &mdash; after that first pass, switching between Remove, Blur, Color pop and Replace, or dragging any slider, re-composites instantly.<\/p>\n<h2>Click any object with MobileSAM<\/h2>\n<p>Auto-detection is great when there&#8217;s one obvious subject. When there isn&#8217;t, <strong>MobileSAM<\/strong> (43&nbsp;MB) lets you simply <em>click<\/em> the exact object you want, and the app traces its true silhouette. That single capability powers three tasks: <strong>Background &rarr; Pick object<\/strong> (cut out or blur one specific thing among many), <strong>Crop &rarr; Select object<\/strong> (snap the crop box to what you clicked), and object-shaped <strong>Blur \/ Mosaic<\/strong> for redaction. No more wrestling a rectangle around an irregular shape.<\/p>\n<h2>Real-ESRGAN: 4&times; upscaling that actually looks sharp<\/h2>\n<p>Resize now has an <strong>enlargement engine<\/strong> choice. Pick <em>Standard<\/em> for an instant scale, or <strong>AI 4&times; (Real-ESRGAN)<\/strong> when you&#8217;re making something bigger and want it crisp &mdash; enlarging a small logo, a low-res product shot, or an old scan. The 67&nbsp;MB model runs <strong>tiled and seam-free<\/strong>, keeps transparency intact, and then fits the result to your exact target size. It&#8217;s the difference between a soft, blocky enlargement and one that holds its edges.<\/p>\n<div style=\"border:1px solid #c5d3f8;background:linear-gradient(135deg,#eef2fd 0%,#ffffff 72%);border-radius:14px;padding:1.5rem 1.65rem;margin:2rem 0;\">\n<div style=\"font-size:.7rem;font-weight:700;letter-spacing:.08em;text-transform:uppercase;color:#2d5be3;margin-bottom:.55rem;\">&#128295; BackendSide Tool<\/div>\n<h4 style=\"margin:0 0 .45rem;font-size:1.15rem;color:#1a1916;font-weight:700;\">ImageLabs 2.0 &mdash; An Offline AI Photo Studio for Windows<\/h4>\n<p style=\"margin:0 0 1.05rem;color:#3d3c38;font-size:.92rem;line-height:1.65;\"><strong>ImageLabs<\/strong> packs nine specialised AI models &mdash; background removal, click-to-select, 4&times; upscaling, colorization and scratch repair &mdash; into one focused Windows app, alongside crop, resize, watermark, OCR, redaction and batch. Every model runs <strong>entirely on your PC<\/strong> with GPU acceleration; no account, no cloud, nothing uploaded. Free to try on the Microsoft Store.<\/p>\n<p>  <a href=\"https:\/\/backendside.com\/imagelabs.php\" style=\"display:inline-flex;align-items:center;gap:.4rem;background:#2d5be3;color:#ffffff;font-weight:600;font-size:.85rem;padding:.6rem 1.2rem;border-radius:6px;text-decoration:none;\">Explore ImageLabs &rarr;<\/a>\n<\/div>\n<h2>Restore old photos: DDColor and LaMa<\/h2>\n<p>A new <strong>RESTORE<\/strong> group brings two of the most satisfying AI tasks to the desktop:<\/p>\n<ul>\n<li><strong>Colorize (DDColor)<\/strong> &mdash; add natural, realistic colour to a black-and-white or faded photo in one click. The original detail (luminance) is preserved at full resolution; only colour is generated. Crucially, it works on <em>true<\/em> greyscale files &mdash; scanned or exported pure B&amp;W images &mdash; not just colour photos that happen to look grey. (935&nbsp;MB model, downloaded on first use.)<\/li>\n<li><strong>Repair \/ Scratch removal (LaMa)<\/strong> &mdash; paint over scratches, tears, creases or small unwanted objects with an adjustable brush, then press Repair. Only the painted areas are filled in; everything else is untouched, with undo-per-stroke and clear. (88&nbsp;MB model.)<\/li>\n<\/ul>\n<p>Together they make a genuine two-step photo restoration: colourise a heirloom print, then paint out the cracks and scratches.<\/p>\n<h2>Beyond the AI: a proper editor around it<\/h2>\n<p>The models sit inside a workspace that grew up in 2.0. The task rail is now grouped into EDIT \/ AI &amp; COLOR \/ RESTORE \/ ANNOTATE \/ TOOLS, with an <strong>Overview<\/strong> screen offering Quick actions, live Image info, and a visible <strong>History<\/strong> of every edit. That history is what powers <strong>Batch<\/strong>: perfect one image (say <em>crop &rarr; AI remove background &rarr; resize<\/em>), then replay the exact chain &mdash; AI steps included &mdash; across a whole folder of product shots.<\/p>\n<p>The everyday tools are all here too, and several are new:<\/p>\n<ul>\n<li><strong>Extract Text (OCR)<\/strong> &mdash; read text out of an image fully offline, with a language picker, an editable box, Copy and Save&nbsp;.txt.<\/li>\n<li><strong>Replace Color<\/strong> &mdash; detected colours become clickable swatches; recolour with a hex value or an eyedropper, including a Transparent plate that erases a colour.<\/li>\n<li><strong>Blur \/ Mosaic<\/strong> for redacting faces, plates and addresses &mdash; by rectangle or by exact object shape.<\/li>\n<li><strong>Add Text<\/strong>, <strong>Auto enhance<\/strong>, <strong>Compress \/ Convert<\/strong> with a live file-size estimate, and a <strong>Metadata<\/strong> viewer with a Strip-EXIF-on-save toggle.<\/li>\n<\/ul>\n<p>Add Redo, keyboard shortcuts, clipboard paste\/copy, hold-to-compare, a split Save button and a modern colour picker on every colour input, and the workflow finally matches the ambition of the AI.<\/p>\n<h2>The full model roster<\/h2>\n<table style=\"width:100%;border-collapse:collapse;margin:.5rem 0 1rem;font-size:.88rem;\">\n<thead>\n<tr style=\"background:#f4f6fb;\">\n<th style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:left;\">Model<\/th>\n<th style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:left;\">What it powers<\/th>\n<th style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">Size<\/th>\n<th style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:left;\">License<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">U&#178;-Netp <em>(bundled)<\/em><\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Background removal &mdash; fast<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">4.7&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Silueta<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Background removal &mdash; balanced<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">43&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">U&#178;-Net<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Background removal &mdash; best quality<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">176&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">U&#178;-Net Human<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Background removal &mdash; people<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">176&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">ISNet<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Background removal &mdash; sharpest edges<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">176&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">MobileSAM<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Click-to-select any object<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">43&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">MIT<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Real-ESRGAN<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">AI 4&times; super-resolution<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">67&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">BSD-3-Clause<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">DDColor<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Colorize B&amp;W \/ faded photos<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">935&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">LaMa<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Repair \/ scratch removal (inpainting)<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;text-align:right;\">88&nbsp;MB<\/td>\n<td style=\"border:1px solid #dce1ec;padding:.5rem .7rem;\">Apache-2.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>Every model is open-source, is integrity-verified (SHA-256) before it&#8217;s ever used, and runs with GPU acceleration and automatic CPU fallback. One is bundled; the rest download on demand the first time you use a feature that needs them &mdash; and an <strong>AI Models<\/strong> manager lets you download, delete, or import a model file on fully offline machines.<\/p>\n<h2>The part that hasn&#8217;t changed: it all stays on your PC<\/h2>\n<p>The headline of 2.0 is the AI, but the principle underneath it is the same one ImageLabs launched with. There is no server doing the heavy lifting. Every model &mdash; segmentation, upscaling, colorization, inpainting &mdash; runs locally, in memory, on your machine. The only time ImageLabs touches the network is the one-time, verified download of an optional model. Your photos never leave your computer, which is exactly how photo editing should work.<\/p>\n<h2>Key takeaways<\/h2>\n<ul>\n<li>ImageLabs 2.0 ships <strong>nine on-device AI models<\/strong>: five background removers (U&#178;-Netp, Silueta, U&#178;-Net, U&#178;-Net Human, ISNet), MobileSAM, Real-ESRGAN, DDColor and LaMa.<\/li>\n<li>The new <strong>Background<\/strong> task adds Blur, Color pop and Replace modes on top of Remove, with cached masks so mode changes are instant.<\/li>\n<li><strong>MobileSAM<\/strong> lets you click the exact object to cut out, crop to, or redact; <strong>Real-ESRGAN<\/strong> adds sharp 4&times; upscaling.<\/li>\n<li>A new <strong>RESTORE<\/strong> group colourises old photos (DDColor) and paints out scratches (LaMa).<\/li>\n<li>OCR, redaction, batch, and a redesigned workspace round it out &mdash; and <strong>every model runs entirely on your PC<\/strong>, nothing uploaded.<\/li>\n<\/ul>\n<div style=\"border:1px solid #c5d3f8;background:linear-gradient(135deg,#eef2fd 0%,#ffffff 72%);border-radius:14px;padding:1.5rem 1.65rem;margin:2rem 0;\">\n<div style=\"font-size:.7rem;font-weight:700;letter-spacing:.08em;text-transform:uppercase;color:#2d5be3;margin-bottom:.55rem;\">&#128295; BackendSide Tool<\/div>\n<h4 style=\"margin:0 0 .45rem;font-size:1.15rem;color:#1a1916;font-weight:700;\">ImageLabs 2.0 &mdash; An Offline AI Photo Studio for Windows<\/h4>\n<p style=\"margin:0 0 1.05rem;color:#3d3c38;font-size:.92rem;line-height:1.65;\"><strong>ImageLabs<\/strong> packs nine specialised AI models &mdash; background removal, click-to-select, 4&times; upscaling, colorization and scratch repair &mdash; into one focused Windows app, alongside crop, resize, watermark, OCR, redaction and batch. Every model runs <strong>entirely on your PC<\/strong> with GPU acceleration; no account, no cloud, nothing uploaded. Free to try on the Microsoft Store.<\/p>\n<p>  <a href=\"https:\/\/backendside.com\/imagelabs.php\" style=\"display:inline-flex;align-items:center;gap:.4rem;background:#2d5be3;color:#ffffff;font-weight:600;font-size:.85rem;padding:.6rem 1.2rem;border-radius:6px;text-decoration:none;\">Explore ImageLabs &rarr;<\/a>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>ImageLabs 2.0 turns a simple Windows photo editor into an offline AI studio \u2014 five background-removal models (U\u00b2-Net, Silueta, ISNet and more), click-to-select with MobileSAM, 4\u00d7 Real-ESRGAN upscaling, DDColor colorization and LaMa scratch repair. Here is everything new, and why every model runs on your own PC.<\/p>\n","protected":false},"author":1,"featured_media":133,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7,2],"tags":[],"class_list":["post-129","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-software-releases","category-windows"],"_links":{"self":[{"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/posts\/129","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/comments?post=129"}],"version-history":[{"count":1,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/posts\/129\/revisions"}],"predecessor-version":[{"id":131,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/posts\/129\/revisions\/131"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/media\/133"}],"wp:attachment":[{"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/media?parent=129"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/categories?post=129"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/backendside.com\/blog\/wp-json\/wp\/v2\/tags?post=129"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}