How to remove an image background for free without credits

Start with the method already on your device

The fastest honest recommendation is not always another website. On a supported Mac, Preview can isolate a subject and remove the background. Apple documents the command under the image-editing controls. If that built-in result handles the hair, hands, product gaps, and shadows correctly, it is the shortest route: no new account, credit balance, or browser model is required.

Microsoft 365 applications also include Picture Format → Remove Background. Microsoft’s workflow adds useful manual controls for marking areas to keep or remove. Use it when the picture is already inside Word, PowerPoint, Excel, or Outlook and you need to correct an automatic mask. Saving the edited picture as a separate file produces a reusable result.

Those methods are not universal. Apple’s command depends on the Mac and software version, while Microsoft’s option assumes access to a supported Office application. A browser tool is useful on Chromebooks, Linux, shared computers, and devices where the built-in cutout is missing or inconvenient.

What “free” means on remove.bg, with receipts

Background-removal pages often use “free” to mean free preview, free trial credit, or free low-resolution export. Those are legitimate offers when the limit is stated clearly, but they are different from an unrestricted full-size download.

The official remove.bg help page, checked on 18 July 2026, says its website is free for personal-use images up to 0.25 megapixels. Its credits documentation gives 625 × 400 as an example of that preview size and says one high-resolution background removal uses one credit. It describes high resolution as supporting up to 50 megapixels. The company offers subscriptions and pay-as-you-go packages; follow the official pages because prices and allowances can change.

That does not make remove.bg a bad product. It provides a mature hosted service, high-resolution paid output, an API, integrations, and editing tools. It may be the better choice for a business that wants automation, support, and predictable centralized processing. The important distinction is that “free preview” is not the same promise as “full-size output at no charge.”

FileGizmo’s live background remover charges no credits and does not reduce the output to a preview tier. It exports a transparent PNG at the processed image dimensions and includes a free manual Restore and Erase brush for correcting the mask. Its tradeoff is local computation: the browser must download and initialize a sizeable model, and the device supplies the memory and processing time.

How to create a transparent PNG locally

1. Open Remove Image Background and choose a JPG, PNG, or WebP file with a clear foreground subject.
2. Select Best edges for products, hair, and difficult outlines, or Fast portraits when speed matters more than the hardest edges.
3. Let the browser load the selected model and process the pixels locally. Keep the tab active, especially on a memory-constrained phone.
4. Inspect the checkerboard preview around hair, fingers, handles, transparent objects, and gaps inside the subject.
5. Open Refine edges. Erase any background remnants, Restore foreground pixels the model removed, and use zoom, pan, undo, and hold-to-compare for detailed checks.
6. Download the PNG and place it over both a light and dark background. The full-resolution result is encoded locally only when you download it.

The first quality-mode run can take noticeably longer because the packaged BiRefNet Lite model must load from FileGizmo’s origin. That network response travels from the website to the device. The selected photograph is passed to a browser worker and is not posted to a FileGizmo image-processing endpoint. Later visits can reuse cached model assets until browser storage is cleared.

Choose the method by outcome, not the headline

Use a built-in Apple action when it is present and the one-click result is clean. Use Microsoft’s background tool when the picture is already in an Office document or the automatic selection needs simple manual keep/remove marks. Use FileGizmo when you want a full-size transparent PNG, manual mask refinement, and no account, credit, or photo upload. Its first local run is heavier because the browser must load the model.

A hosted specialist can still be appropriate for high-volume commercial automation, API integration, manual edge editing, formal support, or devices too weak to run a segmentation model locally. Privacy and price are not the only requirements. Delivery speed, edge quality, repeatability, batch control, colour handling, and the right to use the output commercially can all decide the best tool.

The useful comparison is therefore not “which logo is best?” It is:

• Does the free result preserve the dimensions you need?
• Is a credit, login, watermark, or personal-use restriction attached?
• Does the photograph leave the device?
• Can you correct an imperfect mask?
• Does the output contain a real alpha channel?
• Are the difficult edges acceptable against the final background?

Inspect the cutout where it will actually appear

Automatic segmentation estimates foreground probability pixel by pixel. Fine hair, fur, smoke, glass, lace, motion blur, shadows, and colours shared by foreground and background remain difficult. A cutout that looks convincing on a white checkerboard can reveal a pale halo on black, while a dark fringe can disappear on a dark page and become obvious on white.

Test the PNG against the real slide, storefront, profile background, or design colour. Zoom into hands and gaps inside products. Check that the algorithm did not erase a white sleeve against a white wall or retain a shadow you meant to remove. FileGizmo’s Restore brush paints original pixels back through the mask; Erase makes unwanted areas transparent. Soft edges help blend corrections, while undo and redo make exploratory strokes safe.

Do not confuse background removal with compression. The output uses PNG because JPG cannot store transparent pixels. A PNG can be larger than its photographic JPG source. Resize the cutout to its final display dimensions, then choose an output format that supports transparency if file size matters.

The privacy tradeoff is architectural

A hosted background API receives either the uploaded image or a URL it can fetch. remove.bg’s public API documentation demonstrates multipart image uploads, which is expected for a server API. FileGizmo instead packages the model for browser inference. That eliminates an image upload but transfers the computational cost to your device.

You can verify the difference. Open developer tools, clear the Network panel after model assets finish loading, process a harmless sample, and inspect new requests. A model or worker download has a response body coming to the browser; a photo upload has an outgoing request payload. Never use a passport photo, confidential product image, or private family photograph as the first test file.

“Actually free” should mean the promised output is downloadable without a surprise payment step. It should not mean every method has identical capabilities or zero cost to operate. FileGizmo’s cost is the first model download, local time, memory, and battery. That trade is visible, measurable, and does not require surrendering the photograph.