Remove product photo backgrounds with solid or AI mode
Simple solid backgrounds and complex scenes need different removal methods. Solid mode removes border-connected color, while AI mode uses a model on a device that supports WebGPU.

How to use this tool
- Add a photo and choose solid or AI mode.
- Set solid-color tolerance and the output background.
- Run the task and inspect product edges and transparent areas against the original picture.
- Download the resulting PNG.
Choose the right settings
Adjust the color tolerance for a solid background, or select AI in a WebGPU-capable environment for complex photographs. Choose transparent, white or black output for the destination where it will appear.
Before you start
Prepare a representative source
Start with a copy of the source and decide where the result will be used. A picture for later editing, an upload with fixed dimensions and a printed page have different requirements. Check the supported input and intended output in this guide; a new filename extension alone does not convert the file.
Where size, quality, background or crop controls are available, change one choice at a time on a representative image. Include small lettering and an important edge in your comparison. If the task extracts information rather than changing the picture, decide which source region and values you need to inspect.
Write down the recipient’s requirements and the choices used for a trial, including output type and any size or background setting that applies. If a detail is already blurred, cropped or absent in the original, changing an output setting will not supply the missing evidence. Compare an intermediate copy separately from the original: it may already contain changes from a previous step. This record helps you decide whether to correct the source, revise one choice or use a different supported operation.
Compare the result with its source
Open the downloaded output in the program that will actually use it. For image outputs, compare the same area at the same zoom, inspect all edges and read the saved dimensions and file size. For extracted text, measurements or color values, check sample entries directly against the visible source; a plausible-looking result can still be wrong.
If the output differs from the preview or cannot be created, check the input format, this tool’s limits and the selected settings before changing the source. Repeating compression or enlarging a small image cannot restore detail already lost. Keep the original so another trial starts from the same evidence.
Example input and output
product photo → transparent/solid-background PNGExamples describe the workflow. Results depend on the source file and settings.
Check the saved result before using it
Compare hair, product edges and transparent parts with the source to find missing details. Place a transparent PNG over another background to inspect remaining background pixels and accidental removals.
If the result is different from what you expected
If AI cannot start, check WebGPU support or try solid mode for a suitable source. First use downloads about 219 MB of model data, and difficult edges still require manual review.
Limits to keep in mind
- AI needs WebGPU and a ~219 MB first-use model download. Transparent products and complex edges may need further editing.
Continue with a related tool
This guide explains the public browser tool. Keep a copy of the original and inspect the result before sharing it.