Image Tools May 8, 2026 · 5 min read

Remove Image Backgrounds with AI — Free, Private, No Upload

ToolKnit removes backgrounds with a model running locally in your browser. It downloads processing resources, but does not upload your selected image for segmentation.

Try it now — no sign-up needed

Choose an image, run local background removal and inspect the transparent result before downloading.

Open the tool

How It Works

The tool uses a local segmentation model with WebGPU acceleration where supported and a WebAssembly CPU compatibility mode. Initial setup requests your confirmation before downloading up to 109.5 MB of pinned model and runtime resources from ToolKnit. Switching to CPU may require additional files. The browser can reuse cached resources, but storage eviction or private browsing may require another download.

The model estimates a foreground mask for people, products, animals and other subjects. Hair, transparent surfaces, low contrast and fine edges can need manual correction. Processing time depends on the image, device and selected runtime; accurate edges and a fixed completion time are not guaranteed.

Auto Mode vs. Manual Mode

Auto Mode

Click "Auto Remove Background" to generate an initial cutout. Inspect it before export and refine missed areas when needed. Straightforward starting points include:

  • Portrait photos with clear subjects
  • Product images on simple backgrounds
  • Objects with well-defined edges

Manual Refinement

When the AI doesn't get it perfect — complex hair, transparent objects, or busy backgrounds — switch to manual mode. You get:

  • Brush tool — Paint over areas to mark as foreground (keep) or background (remove).
  • Remove Residue — erase background left in the foreground mask. Restore Subject paints back parts removed by mistake.
  • Adjustable brush size — Fine-tune for detailed edges or broad strokes.
  • Edge softness control — Smooth transitions between foreground and background.
  • Undo — step back through brush edits with Ctrl+Z or Command+Z. Clear Edits removes manual strokes; there is no separate Redo control.

Pro tip: Start with Auto Remove, then switch to manual mode to touch up any problem areas. This combination gives the best results with the least effort.

Step-by-Step Guide

  1. Open the Background Remover
  2. Confirm the first-time resource download and wait for setup. Later visits can reuse available cache; download and initialization times vary.
  3. Drop an image or click to upload (JPG, PNG, or WebP)
  4. Click "Auto Remove Background" and wait for processing to finish
  5. If needed, use the brush tools to refine edges
  6. Click "Download PNG" to save with transparent background

Common Use Cases

  • E-commerce product photos — Clean white or transparent backgrounds for online stores. Amazon, Shopify, eBay listings all look more professional.
  • Social media content — Create stickers, profile pictures, or overlay effects for Instagram, TikTok, and YouTube thumbnails.
  • Presentations — Drop product or team photos onto slides without awkward rectangular borders.
  • Design projects — Extract subjects for collages, composites, or graphic design work.
  • Portrait preparation — create a cutout for a composition. For official ID photos, check the issuing authority's rules; retouching or background replacement may not be accepted.
  • Marketing materials — Flyers, banners, and email headers with clean cutouts.

Supported Formats

  • Input — supported JPG, PNG and WebP files up to 20 MB, with additional device-aware pixel limits
  • Output — PNG or supported WebP can preserve transparency. JPG fills transparent areas with the selected background; transparent preview mode uses white for JPG.

The file-size limit is 20 MB (20 × 1,024 × 1,024 bytes). Device-aware pixel and export-canvas checks also apply. A file below the size limit can still be too large to decode or render on a particular device; reduce dimensions and retry if the tool reports a limit.

What This Local Workflow Offers

Choose a workflow based on its actual upload policy, device requirements and output quality. ToolKnit provides the following local-processing features:

  • No upload — Your image stays on your device. The AI runs locally.
  • No account — No sign-up, no email, no login wall.
  • No watermark — Full-resolution output, no branding overlay.
  • No daily limit — Process as many images as you want.
  • Free forever — No premium tier, no credit system.

The trade-off is local resource use: the initial model download, initialization and image processing consume network data, memory and compute. Available cache can reduce repeat downloads, but it does not eliminate initialization or guarantee offline availability.

Privacy

Your selected image is processed locally rather than uploaded for segmentation. Model resources and setup preferences may remain in browser storage, and downloaded files remain wherever you save them. Closing the tab clears the working session, not those downloads or caches. Ordinary website analytics are separate from image processing.

Ready to remove a background?

Choose an image and create a transparent cutout you can inspect and refine.

Open the tool More guides