Lift cuts any subject out of its background and gives you a clean, transparent PNG in a blink. On-device AI handles the edges, then you can drop in studio backdrops, shadows and crops, or batch a whole folder and export as a ZIP.
Every cutout runs in your browser — the photo is never uploaded, never retained, never trained on. Flip on airplane mode and it still works, so client headshots, product shots and ID photos stay private by default.
Open Lift and drag a photo straight into the page — no signup to try. JPG, PNG and WebP all work. The file loads into your browser and goes no further.
On-device AI isolates the subject using WebGPU, with a WASM fallback on older hardware. The model runs locally, so nothing leaves your machine and it keeps working offline.
Add a studio backdrop, shadow or crop, then download a transparent PNG. Need a set? Batch-process many images at once and grab everything as a single ZIP.
When a headshot, badge photo, or scanned ID needs a clean background for a filing or case file, uploading it to a web cutout service can breach an NDA or client-confidentiality duty. Lift does the cutout on your machine, so the image never leaves your laptop — the same reason teams pick nimbril for redaction. Airplane mode is the proof: it still works with no network.
Drop a whole folder of product shots in, get clean transparent PNGs back, and export them as one ZIP — no per-image upload wait, no monthly credit meter. Built-in crop presets match where the photos are going: 1:1 for Amazon and Etsy, 4:5 for Instagram, 3:4 for Poshmark and Depop. Pair white or studio-gradient backdrops with a soft shadow for a consistent storefront look.
New-hire and team-directory photos arrive in a dozen different settings and lighting. Cut everyone out, drop them on the same studio-grey backdrop, and crop to a uniform frame so the directory looks intentional. Because employee photos are personal data, keeping the processing on-device sidesteps the vendor-data-handling questions an upload tool would raise.
You need one sharp, professional-looking headshot for LinkedIn or a resume and don't want to install software, sign up, or hand a candid photo to a site that keeps it. Open Lift, drop the photo, swap the messy background for a clean white or soft-glow studio backdrop, and download. It's free to try with no account, and your photo stays on your device.
Most background removers send your photo to their servers to run the model, which means a client headshot, an employee photo, or a scan of an ID document briefly lives on someone else's infrastructure under their retention and training terms. Lift never does that — the AI model downloads to your browser once and every cutout runs locally, so the image is never uploaded, stored, or used for training. For anyone handling photos under an NDA, HIPAA, or a confidentiality duty, "never uploaded" isn't a marketing line, it's a compliance fact you can verify by switching to airplane mode and watching the tool keep working.
Lift runs an open AI background-removal model directly in your browser via WebAssembly and WebGPU. On a machine with a supported GPU it uses WebGPU for speed; on other hardware it falls back to a WASM path so it still works. The model downloads once on first use and is then cached, so later cutouts start instantly and need no connection at all.
The model traces a per-pixel foreground mask, so it does well on hair, fur, and fine wisps where a hard cutout would look pasted-on. It's strongest on a clear single subject — a person, a product, a pet — against a reasonably distinct background; very low-contrast subjects or busy backgrounds are where any AI cutout gets less certain. Because everything runs locally, you can iterate freely: try a different crop, swap the backdrop, add or remove a shadow, and re-export as many times as you like without uploading anything or burning credits.
No. Lift removes backgrounds entirely on your device — the AI runs in your browser, so your photos are never uploaded, stored, or used for training. You can confirm it by turning on airplane mode: the tool still works with no connection.
Yes, there's a generous free tier you can use with no signup. Pro ($9/mo or $69/yr) unlocks full-resolution exports, batch processing with ZIP, no watermark, and unlimited history across every nimbril tool. A single-app plan is $4/mo.
Yes. Lift cuts the subject out and exports a clean PNG with a transparent background, ready to drop into a deck, store listing, or design. You can also swap in a studio backdrop, add a shadow, or crop before exporting.
Yes. Batch processing lets you cut out a whole set of images in one pass and download them together as a ZIP — handy for product catalogs, team headshots, or ID photo sets.
Lift uses an AI model to trace the subject's edges, including tricky areas like hair and wisps. It runs on WebGPU when your device supports it for speed, falling back to WASM so it still works on older hardware.
Lift runs an open AI segmentation model locally in your browser. It produces a soft per-pixel mask that handles hair and fine edges well on a clear single subject. The trade-off versus an upload-based service like remove.bg is privacy: the cutout happens on your device instead of on a remote server, so your image never leaves your machine.
Beyond a transparent PNG, you can drop the subject onto solid colors, soft studio gradients (white, grey, warm and cool glows, charcoal), or a sunset gradient, and add a shadow. Crop presets are tuned for where images go — 1:1 for Amazon and Etsy, 4:5 for Instagram, 3:4 for Poshmark and Depop, 9:16 for stories, and 16:9 for banners.
The first time you use Lift, your browser downloads the AI model so it can run on your device. That one-time download is what makes every later cutout fast and fully offline. The model is cached afterward, so repeat use — and batch jobs — start right away with no network.
Yes. Lift uses WebGPU for speed when your device supports it and falls back to a WebAssembly path on other hardware so it still works. It runs best on a desktop or laptop; on a phone it still works but can be slower, especially on the first run while the model downloads.