PixelOkay
Background Remover Pro AI● Local image processing
Tools
PIXELOKAY / IMAGE TOOLS / PRIVATE CREATION

Background Remover Pro AI

Remove simple backgrounds with manual brushes, color selection and shapes—no AI required. Refine a cutout, compose new backgrounds, save a project or export real PNG and JPEG files. Optional local AI can suggest a first mask.

No AI by defaultColor wand + shapesPrivate local edits
WORKSPACE

Background studio No AI · Local

Import → Select/Erase → Refine → Compose → Export. Optional AI is never required.

Live preview

Import a photograph to begin.

Make your subject the star.

Add a portrait, product photo, pet or other image. Your original will stay untouched.

No cutout yet
Choose a photo to get started.

Save your finished image

Transparent mode produces alpha PNG. Other background styles are flattened; JPEG always uses an opaque backdrop. Output is capped at working image dimensions.

Applies to PNG and JPEG; mask PNG always keeps full working dimensions.
HOW IT WORKS

Useful cutouts with no AI, plus optional local segmentation.

01

Remove a simple background without AI

Use a color wand, rectangle or polygon to remove selected regions and refine with brushes. AI segmentation is optional.

02

Perfect the difficult edges

Feather delicate boundaries, tighten color fringes and manually restore or erase mistakes.

03

Export for anywhere

Make a transparent PNG or export a solid, gradient or photographic background design.

Automatic segmentation can miss transparent objects, complex hair and overlapping subjects. Manual corrections and the full-resolution working preview can help, but they cannot reconstruct unseen pixels. Processing resolution is capped for device stability; original metadata is not preserved. The editor does not upload your image; optional AI downloads public model/runtime files only with approval. Work and export resolution is capped, and no tool can reconstruct unseen pixels.

Optional · On-device AI

Assist with a foreground mask

The full manual editor is available without AI. This optional step downloads a compact U²-Netp segmentation model (~4.36 MiB) and a version-pinned ONNX Runtime Web/WASM engine from public hosts, then runs inference in your browser.

Model
U²-Netp · 320 × 320 saliency mask · Apache-2.0
Engine
ONNX Runtime Web 1.30.0 · WASM · MIT
Network
Downloads model/runtime files only; your image is not uploaded
Limitations
Not realistic hair/glass matting; performance varies by CPU and memory

If AI fails or is canceled, your current manual edits remain available. An existing mask is never replaced without confirmation.

Using Background Remover Pro AI

Open a photograph and start with no-AI color selection, rectangle, polygon or a manual brush; use optional AI only if requested. Compare the cutout, add a local background if needed and export the image or mask. Keep the original separately.

AI removal downloads U²-Netp and ONNX Runtime on demand; external providers receive asset requests. Salient-subject segmentation is not specialist hair/glass matting, and difficult edges can need manual correction. Image/working-resolution budgets apply. Editable project JSON includes the working-resolution image, mask and settings, but not full original-resolution pixels. Metadata and camera-resolution preservation are not guaranteed.

References: U²-Netp model · U²-Net attribution and license · ONNX Runtime

Workflow and output choices for Background Remover Pro AI

Start with manual color-wand, rectangle, polygon or brush editing; optional U²-Netp can suggest a foreground mask when requested. Refine the result, compose a background and export an image or editable project.

Transparent PNG preserves the cutout; JPEG flattens the chosen background. Project JSON keeps the working-resolution image, mask and settings, not all original-resolution pixels. Keep the original separately.

Example workflow

  1. Choose a JPEG, PNG or WebP image and begin with a manual selection tool if the background has a simple color or shape.
  2. Erase or restore pixels with brushes and refine the edge; request AI only when you want a suggested subject mask.
  3. Choose a transparent, solid, gradient or local-photo background and compare the export with a saved project.

Authored example: Use a fictional product photo against a flat-color backdrop. Select part of the background with Color wand, inspect the edge and restore a small subject area with the brush. This is an authored workflow, not a segmentation result.

Limits and model handling: Manual tools do not require AI. Choosing AI downloads an approximately 4.6 MB U²-Netp model and ONNX Runtime assets from external providers; inference runs locally and the app does not intentionally send the photo to an inference server. Manual edges still need review.

Background Remover Pro AI questions

Can I remove a background without downloading an AI model?

Yes. Start with the color wand, rectangle, polygon or manual brush. The AI model is only requested if you choose the AI action.

Does optional AI send my image to the model provider?

The model and runtime are downloaded from external providers, but inference runs locally in the browser; the app does not intentionally send your image to an inference server.

Related tools